Content-Based Image Retrieval Features: A Survey

Abstract

Content-Based Image Retrieval (CBIR) systems have been used for the searching of relevant images in various research areas. In CBIR systems features such as shape, texture and color are used. The extraction of features is the main step on which the retrieval results depend. Color features in CBIR are used as in the color histogram, color moments, conventional color correlogram and color histogram. Color space selection is used to represent the information of color of the pixels of the query image. The shape is the basic characteristic of segmented regions of an image. Different methods are introduced for better retrieval using different shape representation techniques; earlier the global shape representations were used but with time moved towards local shape representations. The local shape is more related to the expressing of result instead of the method. Local shape features may be derived from the texture properties and the color derivatives. Texture features have been used for images of documents, segmentation-based recognition,and satellite images. Texture features are used in different CBIR systems along with color, shape, geometrical structure and sift features.

Headnote

ABSTRACT

Content-Based Image Retrieval (CBIR) systems have been used for the searching of relevant images in various research areas. In CBIR systems features such as shape, texture and color are used. The extraction of features is the main step on which the retrieval results depend. Color features in CBIR are used as in the color histogram, color moments, conventional color correlogram and color histogram. Color space selection is used to represent the information of color of the pixels of the query image. The shape is the basic characteristic of segmented regions of an image. Different methods are introduced for better retrieval using different shape representation techniques; earlier the global shape representations were used but with time moved towards local shape representations. The local shape is more related to the expressing of result instead of the method. Local shape features may be derived from the texture properties and the color derivatives. Texture features have been used for images of documents, segmentation-based recognition,and satellite images. Texture features are used in different CBIR systems along with color, shape, geometrical structure and sift features.

Keywords: CBIR, Feature Extraction, Feature Selection, Color, Texture, Shape

Date of Submission: April 06, 2018 Date of Acceptance: April 25, 2018

(ProQuest: ... denotes formulae omitted.)

I.INTRODUCTION:

With more digital applications, the number of images produced is increasing rapidly. Huge databases of images have to be searched to retrieve related images; this is mentioned as CBIR techniques. A manual search of such image databases is not only exhaustive but is also not efficient therefore various approaches have been presented for the improvement of images retrieval. Accurate retrieval of images from a lot of digital images is challenging, therefore, CBIR is considered as an intensive research area [1-3]. Previously used text-based query methods are not as efficient and effective as feature-based methods. The various primitive features such as color, shape, structure, texture, etc can be used to extract the similarities between the images already stored in database and a query by user [4-8].

These features are called low-level features which are utilized in the search for feature extraction images to match to the large image database [9-10]. There are various CBIR systems that use more than one feature. Color is used in combination with both texture and shape for improved results [11-12].

1. Applications of CBIR:

CBIR is being used in various fields like biomedicine [1314], medicine [15], military [16], and commerce [17], Art [18,19], Art Galleries [20-21] and Entertainment like content-basedmultimedia retrieval. For Law enforcement [22], CBIR is used for face-detection [23-24] copydetection [25] while for the control of cyber-crime, CBIR is used in Image-based CAPTCHA [26-27].In the field of bioinformatics, it is used for Automatic Annotation [28], molecular analysis [29], medical diagnosis [30-31] and face recognition [32-54]. The search engines also use CBIR for web-search for images [55]. CBIR is widely being used in biometrics systems. In recent years use of CBIR systems in various fields has been increased.

In this survey paper, a comprehensive overview of the CBIR is presented. In Section 2, CBIR Systems are discussed. Various problem domains of CBIR system are also mentioned along with the classification of CBIR systems. Section 3 gives classification is CBIR systems. In Section 4, few color based techniques used in CBIR systems are discussed. Techniques using the shape featurein CBIR systems are mentioned in Section 5. The summary of various texture based methods over the few years is shown in Section 6.

II.CBIR SYSTEM:

In any CBIR system, there are mainly three phases i.e. feature extraction from query image, feature selection and then the similarity matching. The multi-dimensional feature vectors are created by the CBIR system and these feature vectors are matched with those of database images. The similarity measure or distance measure is done with the help of some algorithm on which score is calculated. The threshold value is pre-defined by the system developer. Those images which have more score than the threshold value are extracted and shown as retrieved results. The Relevance feedback method is adopted so that whether the user is satisfied or not satisfied by the retrieved images then the CBIR system is improved to search keeping in view the selected images by the user.

In Figure 1, a flowchart of CBIR system is shown:

2.1. Image Content Descriptor:

An image content Descriptor can be local or global. It can be specific as well as general. Global uses features of thewhole image and local divides image into parts first. A simple method of partition is to use a division i-e cut image into regions having equal shape and size. They may not be meaningful and significant regions but it is a process to represent global features of any image. Partition the image into similar and homogeneous areas is an improved method with the use of some standard such as the Region Segmentation algorithms. Another complex method is to obtain semantically meaningful objects by Object Segmentation. The image content is further classified into two broad categories as visual content and semantic content.

2.1.1. Visual Content:

The visual content is further classified into two main classes:

2.1.1.1. General Visual Content:

When the features or content of the query image are visible and are generally perceived then they fall into this category. Included common visual contents are the features like shape, texture, color, structure, spatial relationship etc.

2.1.1.2. Domain Specific Visual Content:

When the query is based on such content that requires some domain knowledge then those query image content are domain specific visual content like Human Face Detection needs some prior information about the human facial characteristics. These characteristics are not general; these are specific i-e only related to human facial features.

2.1.2. Semantic Content:

The semantic contentsare either described by textual explanation or by using the complex interpretationmeans which are based on some visual-content.

2.2. Image Retrieval Gaps:

The differences between images stored in database and the query image in retrieval are called gaps. The degree of difference will be how far the two images are i-e the gap between the two images. The gaps are divided into two categories: semantic and sensory.

2.2.1. Semantic Gaps:

The mismatching of information of visual query data and the stored image information in the database is obtained. This selected gape to match the image on the similarities basis is called semantic gap. User entered some queries for which optical likeness does not match completely with human perception. By which a semantic gap between CBIR system and the user is obtained [56]. Semantic retrieval has some limitations. A difficulty present in it is that most of the images have more than one semantic interpretation. Because images used for training have usually short description in form of a caption, therefore, some features might never be recognized. This helps to decrease the amount of images instances used for training and weakens the system's capability to be trained for the concepts that are rare and which have a high variable visual appearance. Semantic retrieval system has a limited vocabulary so it mostly generalizes everything other than the semantic space i.e. for which is not trained [57].

2.2.2. Sensory Gaps:

These are the gaps between the real object and the information in form of computational description obtained from capturing that object in an image form. It is the shortcoming of the image capturing device.

2.3. Problem Domains:

Retrieving images from a varied and huge collection is difficult. Searching images based on the spatial relationship of portions of a query image is a not an easy research problem in CBIR. As segmentation of regions or objects of an image is not feasible, except in few applications. Another common problem in CBIR is a selection of Features.

The divergence between user's query and the result displayed by retrieval system images is called Semantic gap. These semantic gaps are caused due to the low-level similarity between database images and high-level userquery image this result in an error from search algorithms. RF (Relevance Feedback) is the solution for thesame problem in which the user has the authority to accept or reject the result and requests for better similarity in output images.

2.4. CBIR Methods:

There are different methods used in CBIR. Figure 2 represents the most commonly used methods by the CBIR systems for the image retrieval.

The Features extraction is mostly the initial step in any CBIR system. Each feature selected is placed in a vector named Feature Vector. For Feature Extraction some of the techniques used are the Symbol spotting (graphical shape extraction), [58] background removal approach [59] keypoints technique [60]. Some other techniques are the usage of Color, texture, shape, structure, edges as features.

Another method is by the use of Salient region features of the user-defined image. Few examples are the use of the global features of the image [61]. Global features such as the interest points of query image as well as SIFT descriptor are used in CBIR [62-64]. Image Segmentation is also done for better region selection. Common methods are IMALBUM [65], SIMPLIcity, [66] Watershed method and Region growing method (bottom-up).

III. CBIR SYSTEM CLASSIFICATION:

As the CBIR systems are increasing in number there was a need for the classification of these systems. One classification of CBIR systems is shown in Figure 3.Feature such as texture, color, and shape are most frequently used in feature extraction phase of CBIR Systems. The growing need for effective as well as efficient image retrieval has made the selection of an appropriate set of features as the most critical step in the CBIR system formation. In image retrieval research is usually related to the statistical or general methods for the properties of the patches of the image.

IV. COLOR FEATURE:

When viewing an image one of the basic features we observe is color. Color reveals a variety of information. For usage of color as a feature in CBIR, there has been a lot of research done such as the color histogram [68-69], conventional color histogram [70-73], color correlogram, and color moments as shown in Figure 4.

4.1 Color Space:

A point in any 3D color space is represented in form of a pixel. Color spaces like RGB, CMY, and HSV (or HSL, HSB) are widely used. Which color space is the best; there is no consensus for results of images retrieved.

4.1.1 RGB (Red, Green and Blue) Color Space:

In color combination of RGB each pixel has three values which are added together to represent a single pixel value in the color space RGB. Another name for RGB is Additive Primaries as they are added together for giving single pixel value. It is device-dependent and non-uniform.

4.1.2 CMY (Cyan, Magenta and Yellow) Color Space:

CMY consist of three color combinations which are Cyan, Magenta and Yellow. Each pixel has three values which are subtracted to represent a single pixel value in CMY. In CMY color space is formed through absorption of light. It is dependent on the device used and visually non-uniform.

4.1.3 HSV (hue, saturation (illumination) and value of brightness) Color Space:

In computer graphics the most used color space is HSV. The Hue is more suitable for retrieval of an object as it is not affected (invariant) to the deviations in light intensity and direction of the camera. A simple formula [74] can convert the RGB coordinates for an image into HSV coordinates.

4.1.4 Opponent Color Space:

It the three axes obtain as R-G, 2B-R-G, R+G+B. The third axis isolates the information about the brightness which is the advantage of opponent color space. As human is sensitive to brightness, therefore, we require a color space which considers the brightness. Opponent Color Space is not invariant to light intensity and shadows hence it is better than the HSV color space.

4.2 Uniformity:

When a color space is selected, uniformity a most important feature is observed [75]. It means that the viewer observes the two color pairs as the same color with same similarity distance. Thus the distance measured among colors should be directly proportional to the perceptual similarity between them.

4.3 Color Descriptor:

CBIR color descriptor can be local or global. Color space selection is important for the color descriptor. As a color space is selected for the representation of the information of color of a pixel of the query image. Quantization of color-space means those colors which are very close to each other and have small distance should be combined together.

There are certain phrases which are followed when color is used as the descriptor. First of all, a color space is designated. Secondly, the quantization of the color space is completed. After quantization color features are extracted. These color features collectively are known as a color descriptor. An appropriate distance function is derived or any particular CBIR System.

4.4Color Histogram (CH):

An effective and real representation is a color histogram. A pixel may be represented as 3 components in the histogram of color space. For the color histogram, spatial information about histogram is not considered. This simple approach is used to split the sub-areas for histogram calculation [76-78]. The histogram is a statistical global feature which shows distribution intensity. Color histograms are widely used in CBIR. [79]. Color feature is often used in a combination of texture and structure features [80]. Color Histograms are also useful in recognition of objects. In a large number of images, the matching of the histogram may saturate discrimination. Joint Histogram is the solution [81]. Different color histograms methods have been used for image retrieval. Few of the color histogram methods are described below:

4.4.1Conventional Color Histogram (CCH)

It shows the frequency of occurrence of each color (no. of pixels having that particular color) within a given image. Every color is connected to its histogram on basis of color similarity. Dissimilarity basis is not taken into account.

To reduce the number of the histograms, Quantization is done. Quantization means to merge two colors histograms which have a similar color. Thus processing time is lessened. Figure 5 shows the CCH steps.

4.4.2 Invariant Color Histograms (ICH)

Invariant Color Histogram is developed on the basis of color gradients. As the name suggests that it is invariant under any geometry of surface which is locally affine [82].

4.4.3 Fuzzy Color Histogram (FCH):

During merging of different color histograms with the slight difference a problem arises. The color may appear to be similar yet they have a small difference. To overcome this problem FCH concept was introduced. The color space is used for relative greenness-redness and bluenessyellowness) [83].

It also defines the extent of there semblance of the color of a pixel cube for all histograms fuzzy relationship function set given by Hu. It is stronger to quantization complications as the changes in in the density of light. The drawbacks are that it describes only the FCH features and the global color feature of the image is equally high in dimensionality as CCH. There is an extra computation for fuzzy membership function^-.

4.4.4 Color Difference Histogram (CDH):

This is a recent technique for the image color feature representation [84]. Previously the histogram techniques only counted the number of frequency i-e the occurrence of the pixels. CDH takes the total of perceptual difference in uniform colors between any two points of images under various backgrounds along with the edge orientation and color in the color space L·a·b·.

4.4.4.1Block Encoding of Color Histogram for CBIR Applications:

Color is a low-level feature yet important feature which is mostly represented as a histogram in different CBIR systems. There are different coding approaches for histogram so that the space used is less and processing is speedy. Recent research on encoding approaches is done by [85]. They have used the Golomb-Rice coding which codes values of the color histogram in floating point by converting them to integers in some pre-processing steps. The histogram is mostly shown in the sparse matrix.

4.5 Color Coherence Vector (CCV):

In CCV [86] spatial information is included in a histogram in a diverse manner. A histogram bin may be classified into two classes i-e Coherent (used for the uniformlycolored large region) or incoherent (not used for the uniformly-colored large region). CCV of an image can be defined as follows in vector form as in equation - (1):

... (1)

CCV has additional information as compared to Color Histogram, therefore, it gives better retrieval results.

4.6 Color Correlogram (CC):

It shows the association of colors in terms of distance. CC is presented as indexed in a table by sets of color where the probability of occurrence of any pixel of color j at a distance d from a pixel of I in the given image is represented by the entry for row (i, j) [87].

The space between pixels P1 and P2 is given by in equation - (2):

... (2)

The d with a small value is enough to show special correlations as local correlations are more important than global correlation.

CC shows local as well as global spatial information. The disadvantage of CC is that it has high-dimensionality of feature space. For reduction of this dimensionality from (O(N2d)) to (O(Nd)) a new feature was introduced. It is known as auto-color correlogram. Auto-correlogram confines the spatial correlation of identical levels only which is shown as given in equation - (3)

... (3)

4.7Color Moment (CM):

CMare using frequently in CBIR systems which are based on color features. CM of First Order is normally mean, Second Order is Variance and the third order is mainly skewness. These are used in representing color distributions.

4.7.1 Mean - First Order Color Moment:

It is defined by the equation-(4):

... (4)

4.7.2 Variance - Second Order Color Moment:

Second order color moment is given by the equation - (5) given below:

... (5)

4.7.3 Skewness - Third Order Color Moment:

Color moment third order (skewness) is given in equation (6): i

... (6)

The value of the ith-color component of any image pixel j is given by fij and the total no. of pixels of an image are represented by N. Color Moment is a compact feature as only three moments for each of the three color component are required i-e a total of nine values is needed. Although it is compact yet there is a problem of lower discrimination power.

4.8Invariant Color Features:

Color changes with the change in illumination [88]. Variability is considered [89]. Invariance to such environmental factor is not regarded as part of most color feature [90-91] Equation for the invariant color feature is as follows given in equation -(7):

... (7)

It only depends on the surface and the sensor. For shiny surfaces its equation is given as equation-(8):

... (8)

This color-shaped based method considers the color, area and a new feature called perimeter-intercepted lengths of the object-segmentation in a given image [92].The image is segmented into K clusters with the help of K-means algorithm. The object's shape is distinguished by the PILs (perimeter-intercepted lengths). PILs are acquired by the interception of 8 line-segments from the object perimeter. All these 8 lines have different orientations yet pass through the center of the object [93].

4.10 Color Image Retrieval Technique based on Color Feature:

From a database of images, for retrieval of similar images, an effective and efficient technique was used by [103]. Some color distributions are used with standard and mean deviation for representation of the global features in any image. The local characteristics of the image were represented by a bitmap of the image. This local characterization increased the accuracy of the image retrieval system.

V.SHAPE FEATURE:

A basic aspect of segmented regions of an image is its shape feature. Its representation is important in the retrieval of similar images [11]. The shape characteristics are extracted and matched with the database image collection, therefore; the shape feature representation is most significant [10].

There are different methods introduced to help in better retrieval by using different shape representation techniques [104]. Earlier the global shape representations were used but with time they move towards local shape representations as local descriptors produce better retrieval results [105-106]. Shape representation using discrete curve evolution for contours simplification removes irrelevant shape features [107].

5.1. Shape Context (Shape Descriptor):

For similarity matching a shape descriptor called shape context was proposed which was useful for varying geometric transformations [108]. Shapes approximation as concave and convex segments was the approach of Dynamic Programming for shape matching. The only problem with it was that there was pre-computation for the Fourier Descriptor and moments in this technique which was slow [109].

5.2. Local Shape Features:

In the local shape feature, all the characteristics related to the geometric details of an image [110]. The local shape is more related to the expressing of result instead of the method [111]. Local shape features may be derived from the texture properties and the color derivatives.

5.3. Scale-Space Theory for Shape Feature:

Scale-Space theory was advised for capturing all the prominent information, [112]. It gives the basis for the detection and recognition of some prominent features and details on any scale.

5.4. Shape Affine-Invariant Descriptor:

Differential geometric invariant and affine-invariant descriptor are also available [113]. Color feature and shape have been combined in invariant method [114-115].

5.5. Object-Shape Feature:

Theoretically, the best approach by segmenting the object for the object-specific information of imagesis within the image [116]. It is not necessary to recognize the exact location of features of the object in an image but it is important to identify the presence of an object with the help of its properties. When segmentation is done as shown in the equation-(9)

... (9)

The shape of an object is shown in equation-(10):

... (10)

Where h shows the functional computing shape and X show the aggregation operation. This equation is for the accumulative features. Some features have general applicability like moment invariants and the Fourier features of objects [117].

For the image retrieval, there is a requirement for the shape representation that measures the distances of deformations. Some mismatch images are acceptable in certain interactive use of retrieval. So compromise on accuracy may be possible as we need a system to be robust and computationally efficient [118]. There are sophisticated methods like elastic deformation of image portions, multi-resolution representation of shapes and modal matching techniques [119-120].

5.6.Contour Multi-scale Model:

Multi-scale models of contours have been studies for database image representation. There are two approaches to Contour multi-scale model:

a) Multi-Scale Model of Contours as Database Image Representation

b) Multi-resolution Shape Model for Contours.

These two approaches for contours use are shown in Fig 6.

5.7.Contour Representation:

The images have to process for retrieval which comprises a group of operations on the contours. Edge-contours are represented by the B-spline. There are three detection types of symmetries: Skew and parallel symmetries are used for inferring shape from the contour. The third type of symmetry is smooth local symmetries used for planar shape description [121].

VI.TEXTURE FEATURE:

In Image processing as well as in computer graphics and computer vision the texture features are widely used [122] like wavelet transforms, discrete cosine transform [123] and curvelet transform. Many textures are composed of the small textons which are numerous and thus cannot be perceived as separate objects.

6.1 Structural Pattern Feature:

Various texture features are proposed for analyzing the structural pattern information in an image. These texture features include Co-Occurrence of gray-scale values [124], multi-resolution analysis [125]; Local Pattern Spectrum [96] and texture unit spectrum-based [126-127] techniques.

6.2 Gabor-Filter Based Techniques for Texture Feature:

Popular Techniques include Gabor-filter based techniques [128-129] Wavelet Decomposition with energy signatures [130] and Gaussian Markov random field (GMRF) models. The query image is afterward convolved with each Gabor functions obtained after the parameter settings. It has been shown that for the highest texture retrieval results, Gabor Transform for texture feature is used in CBIR [131]. Results of the Gabor Transform function is an over-complete demonstration of the user's query image with a certain redundant ratio of K·S (scales·orientation).

6.3Modified Gabor Function:

Gabor transform has to be applied from all orientations as it depends on the direction. The angles used have an effect on the results of the image retrieval. Therefore the existing Gabor transform had to be modified. The modified Gabor transform is independent of the angle selection. Its performance is almost equal to the simple Gabor Transform.

The modified Gabor Transform is invariant to the rotation. Because of this invariance for rotation, the modified Gabor Transform showed better results in almost all the images [132].

6.3.1Contourlet Transform for Texture Features:

These are based on directional multi scale filter bank which deals with the smooth contours of the images. Contourlets allow various directions at every scale which in turns gives a critical sampling. They give directionality as well as wavelets features. Figure 7 expresses the comparisons of Gabor Wavelet Transform Method and Contourlet Transform approach is shown.

6.3.2Gray-level Run Length Technique for Texture Feature:

Gray-level run length is useful when analysis at more than single orientation is to be done for the extraction of structural information. Spectrum-based techniques are also not recent [133]. With time signal-processing has attained progress which results in the efficient power spectrum tools for analysis. Earlier techniques were gray-level run length [134] gray-level co-occurrence matrix [122]. Some models utilize statistical regularities in the texture field while other models take texture as result of a deterministic dynamic system which observes noise.

6.4Techniques for Transformations of Texture Surfaces:

A collective problem is that for the texture feature extraction the image with texture surfaces has to undergo numerous scaling and re-orientation steps but it is a computationally difficult task. Previously Pyramid Structured Wavelets [135] tree-structured wavelets [136]; multi-resolution simultaneous autoregressive models were the multi-resolution technique but the Gabor filters have shown better performance [137].

6.4.1Techniques for Illumination Effect on Texture Surfaces:

The recent algorithms proposed for the texture analysis consider the non-ideal conditions like illumination intensity change due to the source itself or the capturing device, texture viewing angle change and spectral attributes of light [138]. In the real-world scenario, there are certain environmental factors like changes of light directed from source [139]. Changes in the spectral attributes of light also have an effect on the texture of the image surface affecting the inter-class distances as well as the intra-class of the textures [140].

6.4.2Algorithms for Changes in Affine Transformation and Light Attribute:

Affine Transformations like rotation and scaling of images affect the texture surfaces [141-142], the non-affine transformations like 3D orientation changes also affect texture characteristics [143]. In the real-world, the 3D nature of the textured surface may significantly change the observed texture data by light direction change [144]. Some of the recent texture analysis algorithms are LPB uniform for the changes in spectral attributes of light and LPBrotational invariance for the changes in affine transformations [145].

The common texture analysis algorithms are Gray-Level Co-occurrence matrix (GLCM), Gabor filters, LBP with Rotation Invariance, Local Binary Patterns (LBP), LBP with Contrast Measures, LM 3-D Textons [146], MR8Textons, Multi-scale Opponent Color Representation and Multi-resolution Histogram [147]. In graphics field, there are researches in the texture synthesis field where on the adjacent orientation of pairs of wavelet coefficients in the images, the statistical descriptor is based. To extract the texture features the coefficients of a transform are used on the original pixel values of an image by statistical computations [148].

6.5Texture Descriptor:

Texture feature was used for image retrieval even in 1996[149]. Texture Descriptors encoded general visual attributes into a numerical form which can be easily used for the computational tasks [150]. Some advance textured descriptors are affine transform and photometric transformation invariant features [151]. Improved affineinvariant transform designed for the texture recognition were used in point detection [152].

6.5.1 Wavelet Transform:

Wavelet feature performs well when the texture images were rotated [153]. Wavelets have been used for their compression efficiency and their locality. The various wavelet transforms generates series of rotation and dilations. Lowest levels of wavelet transforms are used for the texture representation [154-156]. A comparative study of the early works done on texture classification from transform-based characteristics is done [157].

6.5.2 Wavelet Packet Signature Methodology:

A generalization of supported wavelets is Wavelet packets [158]. The use of wavelet packet was initiated by Coifman et al.[159].The wavelet packets are ortho normal [160] and are localized in both frequency and time and thus can be used an alternate for the Fourier (frequency based) analysis. The methodology for the texture classification by the help of wavelet packet representation was introduced [154]. The minimum-distance classifier used was Euclidean Distance. Texture Classification by Wavelet Packets Signature Methodology is shown in Figure 8.

6.5.3 Multi-Resolution Wavelet Packet Method:

Texture signatures that were based on the multi-resolution wavelet packet method showed potential for accurate classification and discrimination. For texture segmentation, the methods of scale-space analysis which was done using this method gave efficient representation [161].

6.5.4 Wavelet Packets and Fuzzy Spatial Relations:

In CBIR system the relevant database images with respect to the user's defined image are recovered using the features like texture, color, and shape [162]. But there was a need for more sophisticated image retrieval system for this a method based on the wavelet packets was introduced. The Fuzzy similarity measure was introduced in FIRM [163] in which fuzzy features were used. Wavelet transform provides structure for analysis and at different scales properties of images [164].

6.5.5 Local Spatial Frequency Properties for Texture Features:

Local spatial frequency properties can be utilized to obtain attributes of texture. These attributes can be shared in the wavelet area by wavelet filter for convolution in the spatial area. For the extraction of colored texture mixing of wavelet features was done [165]. For multi-resolution Feature Extraction, the arbitrary tree structure was used, analysis of multi-resolution considers the spatial similarities across sub-bands [166].

6.5.6 Application of Texture Features:

Texture features have been used for images of documents, [167] segmentation-based recognition, [168] and satellite images. [169] Texture features are used in various CBIR systems along with color, shape, geometrical structure and sift features.

6.6 Color Co-Occurrence Matrix for Texture Feature Extraction:

A unique method is introduced which is used to color cooccurrence for extraction of texture feature and for measuring similarity into two color images [170]. The feature obtained gives information about the texture correlation [171] and color information. This technique gave better result in terms of retrieval accuracy (recall and precision) as compare to the GLCM and histogram technique.

Classification of various methods used for feature extraction mostly comprises of two main categories statistical based feature extraction and structural based feature extraction. Statistical feature extraction methods are a morphological operator, adjacency graph, and cooccurrence matrices while statistical feature extraction methods are world decomposition and Markov random field. The categories are shown by the help of following diagram [172-173]. Categories of Feature Extraction Methods are given in Figure 9.

VII.CONCLUSION:

Large databases of images have to be searched to retrieve query related images with the help of CBIR System. A manual search of such image databases is inefficient and exhaustive. Accurate retrieval of images from numerous digital images is challenging, therefore, CBIR Systems is an intensive research area. Semantic Gap is a deviation from the information that is acquired from visual query data and the already stored information in the database on the similarities basis. For better retrieval of images from databases, certain features are used. Feature such as shape, color and texture are most generally used in feature extraction phase. Growing need for effective image retrieval has made the feature selection phase the most critical step. This survey paper consists of the brief introduction of features of CBIR system. Color feature is observed at first. Different color spaces are defined in research work related to the use of the color feature in CBIR system. A comparison of different techniques using color as the main feature is presented in this survey. The shape is the basic characteristic of segmented regions of an image. The features of shape are extracted and matched with features already saved in database for images. Global shape representations, as well as local shape representations, are used forshape descriptors produce better retrieval results. Local shape descriptors are related to the geometric details of an image. Local shape features may be derived from the texture properties and the color derivatives. Texture features are used in various CBIR systems along with color, shape, geometrical structure and sift features. Some advance textured descriptors are affine transform and photometric transformation invariant features. Improved affine-invariant transform designed for the texture recognition were used in point detection. A common problem with texture feature extraction is that the image with texture surfaces has to undergo numerous scaling and re-orientation steps which are computationally difficult tasks. Previously Pyramid Structured Wavelets, tree-structured wavelets, multi-resolution simultaneous autoregressive models were the multi-resolution technique but the Gabor filters have shown better performance. Some recent work in the field of CBIR features is also discussed in the survey paper.

Sidebar

AUTHORS' PROFILES

Anum Masood, Lecturer, Department of Computer Science, CIIT, Wah Campus, received her Master of Computer Science and MS (CS) degree from COMSATS Institute of Information Technology, WahCantt. She is now serving as Lecturer, Department of Computer Science, CIIT, Wah Campus. Her areas of interest are Image Processing, Computer vision, and Medical Imaging.

Muhammad Alyas Shahid received his Master of Computer Science degree in 2002. He received his MS(CS) in 2017 from COMSATS Institute of Information Technology, WahCantt with specialization in Image Processing. He is into teaching field from 1998 till date. Currently he is serving as a Lecturer of Computer Science in POF Institute of Technology, WahCantt. His research interests are Image Processing, Multimedia Processing, and Computer Networks & Security.

Muhammad Sharif, PhD, Associate Professor COMSATS Institute of Information Technology, WahCantt received his MSc in Computer Science from Quaid-e-Azam University, Islamabad. He received his MS(CS) and PhD(CS) from COMSATS Islamabad with specialization in Image Processing. He is into teaching field from 1995 till date. His research interests are Image Processing, Computer Networks & Security, Parallel and Distributed Computing (Cluster Computing) and Algorithms Design and Analysis.

References

REFERENCES

[1] Smeulders, A.W., M. Worring, S. Santini, A. Gupta, and R. Jain. "Content-Based Image Retrieval at the end of early years." IEEE Trans. Pattern Analysis and Machine Intelligence 22, no. 12 (2000): 1349-1380.

[2] Enser, P.G.B., "Pictorial information retrieval", J. Document 51, no. 2 (1995): 126-170.

[3] Liu, Y., Zhang, D., Lu,G., Ma, W., "A survey of content-based image retrieval with high semantics". Pattern Recognition 40, no.1 (2007): 262-282

[4] Remco, C., Veltkamp, and MirelaTanase,"Content-based image retrieval systems: A survey". Technical Report UU-CS2000-34 (2001).

[5] Liu, Ce, Jenny Yuen, and Antonio Torralba. "SIFT Flow: Dense Correspondence Across Scenes and Its Applications." In Dense Image Correspondences for Computer Vision, pp. 1549. Springer International Publishing, 2016.

[6] Babu, R. Bulli, V. Vanitha, and K. SaiAnish. "CBIR using Color, Texture, Shape and Active Re-Ranking Method." Indian Journal of Science and Technology 9, no. 17 (2016).

[7] Sangkloy, Patsorn, Nathan Burnell, Cusuh Ham, and James Hays. "The sketchy database: learning to retrieve badly drawn bunnies." ACM Transactions on Graphics (TOG) 35, no. 4 (2016): 119.

[8] Liu, P., Jia, K.,Wang, Z., and Lv, Z. "A New and effective Image Retrieval Method based on Combined Features", Proc. IEEE Int. Conf. on Image and Graphics '07, (August 2007):786790.

[9] Gevers, Theo, and Arnold WM Smeulders. "Pictoseek: Combining color and shape invariant features for image retrieval." Image Processing, IEEE Transactions on 9.1 (2000): 102-119.

[10] Zhang, Jianlin, and WenshengZou. "ContentBased Image Retrieval using color and edge direction features." Advanced Computer Control (ICACC), 2010 2nd International Conference on. Vol. 5. IEEE (2010).

[11] Saha, S.K., Amit, K.D., and Bhabatosh, C. "CBIR using perception based texture and color measures." Pattern Recognition, 2004. ICPR 2004. Proceedings of the 17th International Conference on. Vol. 2. IEEE, (2004).

[12] Yang, Mingqiang, KidiyoKpalma, and Joseph Ronsin. "A survey of shape feature extraction techniques." Pattern recognition (2008): 43-90.

[13] Ma, Liguang, Yanrong Cao, and Jianbang He. "Biomedical Image Storage, Retrieval and Visualization Based-on Open Source Project." Conference on Image and Signal Processing CISP '08. 2008. 63- 66 .

[14] Shapiro, L. G., I. Atmosukarto, H. Cho, H. J. Lin, S. Ruiz-Correa, and J. Yuen. "SimilarityBased Retrieval for Biomedical Applications." Case-Based Reasoning on Images and Signals Studies in Computational Intelligence. 2008. 355-387.

[15] Muller, H., N. Michoux, D. Bandon, and A. Geissbuhler. "A review of content-based image retrieval systems in medical applicationsClinical benefits and future directions." International Journal of Medical Information 11 (2004): 73.

[16] Gudivada, V.N., and V.V. Raghavan. "CBIR systems." Computer 28, no. 9 (1995): 18-22.

[17] Ruia, Yong, Thomas S. Huang, and Shih-Fu Chang. "Image Retrieval: Current Techniques, Promising Directions, and Open Issues." Journal of Visual Communication and Image Representation 10, no. 1 (March 1999): 39-62.

[18] Goodall, Simon, et al. "SCULPTEUR: Multimedia Retrieval for Museums." Image and Video Retrieval-Lecture Notes in Computer Science (Springer) 3115 (2004): 638-646.

[19] Allen, P.J., R. Vaccaro, and G. Presutti. "ARTISTE: An Integrated Art Analysis and Navigation Environment." Cultivate Interactive 1 (2000).

[20] Kekre, H.B., and Sudeep D. Thepade. "Improving 'Color to Gray And Black' using Kekre's LUV Color Space." IEEE International Advanced Computing Conference. Patiala, India, 2009.

[21] Thepade, Sudeep D., and H.B. Kekre. "Color Traits Transfer to Grayscale Images." Proceedings Of IEEE First International Advanced Computing Conference 2010. 2010.

[22] Goodrum, Abby A. "Image information retrieval: An overview of current research." Informing Science 3, no. 2 (2000): 63-66.

[23] Erik Hjelmås, 1, Boon Kee Low. "Face Detection: A Survey." Computer Vision and Image Understanding 83, no. 3 (2001): 236274.

[24] Abdel-Mottaleb, M., A.K. Jain, and Rein-Lien Hsu. "Face detection in color images." Pattern Analysis and Machine Intelligence, IEEE Transactions 24, no. 5 (May 2002): 696-706.

[25] Lee, Jung-Eun, Rong Jin, N. Gregg, and A.K. Jain. "Content-based image retrieval: An application to tattoo images." 16th IEEE International Conference on Image Processing (ICIP). 2009. 2745- 2748]

[26] Von Ahn, Luis. "CAPTCHA: Using hard AI problems for security." Advances in Cryptology-EUROCRYPT, 2003: 646-646.

[27] Datta, Ritendra, Jia Li, and James Z. Wang. "IMAGINATION: a robust image-based CAPTCHA generation system." Proceedings of the 13th annual ACM international conference on Multimedia. ACM, 2005.

[28] Wenyin, L., S. Dumais, Y. Sun, H. Zhang, M. Czerwinski, and B. Field. "Semi-automatic image annotation." Proceedings of interact: Conference on HCI. 2001. 326-333.

[29] Freitas, M. P., Brown, S. D., & Martins, J. A. " MIA-QSAR: a simple 2D image-based approach for quantitative structure-activity relationship analysis." Journal of molecular structure 738, no. 1 (2005): 149-154.

[30] I. Irum, M. Raza, and M. Sharif, "Morphological Techniques for Medical Images: A Review," Research Journal of Applied Sciences, Engineering and Technology, Vol. 4, pp. 2948-2962, 2012.

[31] Yasmin, Mussarat, Muhammad Sharif, and SajjadMohsin. "Neural Networks in Medical Imaging Applications: A Survey", World Applied Sciences Journal 22, no. 1 (2013): 8596.

[32] Muhammad Sharif, Adeel Khalid, MudassarRaza, SajjadMohsin, "Face Recognition Using Gabor Filters", Journal of Applied Computer Science & Mathematics, Volume 5, issue 11, pp 53-57, 2011

[33] Muhammad Sharif, SajjadMohsin, Muhammad YounasJaved, and Muhammad Atif Ali," Single Image Face Recognition Using Laplacian of Gaussian and Discrete Cosine Transforms" the International Arab Journal of Information Technology (IAJIT), 9, no. 6 (2012): 562-570.

[34] Aisha Azeem, Muhammad Sharif, MudassarRaza, MarryamMurtaza, "A Survey: Face Recognition Techniques Under Partial Occlusion", The International Arab Journal of Information Technology (IAJIT) Volume 11, No. 1, January 2014

[35] Muhammad Sharif, SajjadMohsin, Muhammad Jawad Jamal, MudassarRaza, " Illumination Normalization Preprocessing for face recognition", in Environmental Science and Information Application Technology (ESIAT), 2010 International Conference on, 2010, pp. 4447

[36] Muhammad Sharif, Jamal H. Shah, SajjadMohsin, and MudassarRaza,"SubHolistic Hidden Markov Model for Face Recognition", Research Journal of Recent Sciences, Vol. 2, pp. 10-14, 2013.

[37] Muhammad Sharif, Muhammad Atif Ali, MudassarRaza, SajjadMohsin, "Face Recognition using Edge Information and DCT", Sindh University Research Journal (SURJ) (SCIENCE SERIES), Volume 43 (2) 209-214 (2011)

[38] MarryamMurtaza, Muhammad Sharif, MudassarRaza, Jamal Hussain Shah, "Analysis of Face Recognition under Varying Facial Expression: A Survey", The International Arab Journal of Information Technology (IAJIT) Volume 10, No.4 , July 2013

[39] Jamal Shah, Muhammad Sharif, MudassarRaza, Aisha Azeem, "A Survey: Linear and Nonlinear PCA Based Face Recognition Techniques", The International Arab Journal of Information Technology (IAJIT) Volume 10, No. 6, November 2013

[40] Sharif, Muhammad, SaadAnis, MudassarRaza, and SajjadMohsin. "Enhanced SVD Based Face Recognition." Journal of Applied Computer Science & Mathematics 12 (2012).

[41] Muhammad Sharif, Muhammad YounasJaved, SajjadMohsin, "Face Recognition Based on Facial Features", Research Journal of Applied Sciences, Engineering and Technology 4(17): 2879-2886, 2012

[42] Murtaza, Marryam, Muhammad Sharif, MudassarRaza, and J. Shah. "Face Recognition using Adaptive Margin Fisher's Criterion and Linear Discriminant Analysis." International Arab Journal of Information Technology 11, no. 2 (2014): 1-11.

[43] Muhammad Sharif, SajjadMohsin, Muhammad Jawad Jamal, Muhammad YounasJaved, MudassarRaza, "Face Recognition for Disguised Variations Using Gabor Feature Extraction" Australian Journal of Basic and Applied Sciences, Vol. 5, pp. 1648-1656, 2011.

[44] Muhammad Sharif, Kamran Ayub, Danish Sattar, MudassarRaza, SajjadMohsin," Enhanced and Fast Face Recognition by Proposing Hashing Algorithm", Journal of Applied Research and Technology (JART), Vol. 10, pp. 607-617, 2012.

[45] Muhammad Sharif, Muhammad Kamran Ayub, MudassarRaza and SajjadMohsin," Data Reductionality Technique for Face Recognition", Proceedings of the Pakistan Academy of Sciences, 48 (4): 229-234, 2011

[46] Muhammad Sharif, SajjadMohsin, Rana Abdul Hanan, Muhammad YounasJaved, MudassarRaza, "Using Nose Heuristics for Efficient Face Recognition", Sindh University Research Journal (SURJ) (SCIENCE SERIES) Vol.43 (1-A) 63-68 (2011)

[47] Shah, Jamal Hussain, Muhammad Sharif, MudassarRaza, and Aisha Azeem. "Face recognition across pose variation and the 3S problem." Turkish Journal of Electrical Engineering and Computer Science 22, no. 6 (2014): 1423-1436.

[48] Muhammad Sharif, Adeel Khalid, MudassarRaza, SajjadMohsin, "Face Detection and Recognition Through Hexagonal Image Processing", Sindh Univ. Res. Jour. (Sci. Ser.) Vol.44 (3) 359-366 (2012)

[49] Shah, Jamal Hussain, Muhammad Sharif, MudassarRaza, and MarryamMurtaza. "Robust Face Recognition Technique under Varying Illumination." Journal of Applied Research and Technology 13, no. 1 (2015): 97-105.

[50] Muhammad Sharif, SajjadMohsin, Rana Abdul Hanan, Muhammad YounasJaved, MudassarRaza, "3d Face Recognition Using Horizontal And Vertical Marked Strips", Sindh University Research Journal (SURJ) 2011 Volume 43 No.01-A JUNE http://www.usindh.edu.pk/surj/261-vol-43-1a2011

[51] Muhammad Sharif, SajjadMohsinand Muhammad YounasJaved," A Survey: Face Recognition Techniques", Research Journal of Applied Sciences, Engineering and Technology, Vol. 4. No. 23 pp 4979-4990, 2012

[52] Sharif M., Mohsin S., Jamal M. J. and Raza M., "Illumination Normalization Preprocessing for face recognition", IEEE International Conference on Environmental Science and Information Application Technology (ESIAT), , 44-47 (2010)

[53] Sharif M., Ayub K., Sattar D. and RAZA M., "Real Time Face Detection", Sindh Univ. Res. Jour. (Sci. Ser.) Vol. 44(4), 597-600,(2012)

[54] Sharif, Muhammad, Farah Naz, MussaratYasmin, Muhammad AlyasShahid, and AmjadRehman. "Face Recognition: A Survey." Journal of Engineering Science & Technology Review 10, no. 2 (2017).

[55] Kherfi, M. L., D. Ziou, and A. Bernardi. "Image Retrieval from the World Wide Web: Issues, Techniques, and Systems." ACM Computing Surveys 36, no. 1 (March 2004): 35-67.

[56] Vasconcelos, Nuno. "From pixels to semantic spaces: Advances in content-based image retrieval." IEEE Computer Society, 2007: 2026.

[57] Rasiwasia, N., N. Vasconcelos, and P.J. Moreno. "Query by Semantic Example." Proceedings 5th International Conference on Image and Video Retrieval. LNCS 4071, Springer, 2006. 51-60.

[58] Tombre, K., and B. Lamiroy. "Graphics recognition - from re-engineering to retrieval." Proceedings on Seventh International Conference on Document Analysis and Recognition. 2003. 148-155.

[59] Elhabian, S. Y., K. M. El-Sayed, and S. H. Ahmed. " Moving object detection in spatial domain using background removal techniquesstate-of-art." Recent patents on Computer Science 1, no. 1 (2008): 32-54.

[60] Rusinol, M., and J. Llados. " Logo spotting by a bag-of-words approach for document categorization." 10th International Conference on In Document Analysis and Recognition, ICDAR'09. IEEE, 2009. 111-115.

[61] Shao, H., T. Svoboda1, T. Tuytelaars, and L.e. Van Gool. " HPAT indexing for fast object/scene recognition based on local appearance ." Image and Video Retrieva, 2003: 307-312.

[62] Hare, J. S., P. H. Lewis, P. G. Enser, and C. J. Sandom. "Mind the Gap: Another look at the problem of the semantic gap in image retrieval." 2006.

[63] Sebe, N., Q. Tian, E. Loupias, M. S. Lew, and T. S. Huang. "Evaluation of salient point techniques." Image and Vision Computing 21, no. 13 (2003): 1087-1095.

[64] Lowe, D. G. " Object recognition from local scale-invariant features." The Proceedings of the Seventh IEEE International Conference on Computer Vision. (IEEE) 2 (1999): 1150-1157. [77]

[65] Idrissi, K., G. Lavoué, J. Ricard, and A. Baskurt. "Object of interest-based visual navigation, retrieval, and semantic content identification system." Computer Vision and Image Understanding 94, no. 1 (2004): 271- [78] 294.

[66] Wang, J. Z., J. Li, and G. Wiederhold. "SIMPLIcity: Semantics-sensitive integrated [79] matching for picture libraries." Pattern Analysis and Machine Intelligence (IEEE) 23, no. 9 (2001): [80] 947-963.

[67] Hoiem, D., R. Sukthankar, H. Schneiderman, and L. Huston. "Object-based image retrieval using the statistical structure of imagesn." Proceedings of IEEE Computer Society Conference In Computer [81] Vision and Pattern Recognition, 2004. CVPR'04 . IEEE, 2004. 490.

[68] Swain, M.J., and D.H Ballard. "Color Indexing." International Journal of Computerl Vision 7, no. 1 (1991): 11-32.

[69] Yasmin, Mussarat, SajjadMohsin, IsmaIrum, and Muhammad Sharif. "Content based image retrieval by shape, color and relevance feedback." Life Science Journal 10, no. 4s (2013): 593-598.

[70] MussaratYasmin, Muhammad Sharif, and Muhammad AlyasShahid. "Content Based Image Retrieval Based on Color: A Survey." International Journal of Advanced Networking and Applications 7, no. 3 (2015): 2724.

[71] MussaratYasmin, Muhammad Sharif, IsmaIrum, WaqarMehmood, and Steven Lawrence Fernandes. "Combining Multiple Color And Shape Features For Image Retrieval."

[72] MussaratYasmin, Sharif Muhamma, MohsinSajjad, and IrumIsma. "Content Based Image Retrieval Using Combined Features of Shape, Color and Relevance Feedback." KSII Transactions on Internet & Information Systems 7, no. 12 (2013).

[73] MussaratYasmin, Muhammad Sharif, IsmaIrum, and SajjadMohsin. "An efficient content based image retrieval using EI classification and color features." Journal of applied research and technology 12, no. 5 (2014): 877-885.

[74] Foley, J.D., A.Van Dam, S.K. Feiner, and J.F. Hughes. Computer Graphics: Principles and Practice. 2nd. Addison-Wesley, 1990.

[75] Mathias, E. "Comparing the influence of color spaces and metrics in content-based image retrieval." Proceedings of International Symposium on Computer Graphics, Image Processing and Vision. 1998. 371-378.

[76] Flickner, M. et al.,. "Query by image and video content:The QBIC system." IEEE Computer Vision 28, no. 9 (1995): 23-32. [48]Deng, Y., B.S. Manjunath, C. Kenney, M.S. Moore, and H.D. Shin. "An efficient color representation for image retrieval." IEEE Transaction Image Processing 10, no. 1 (2001): 140-147.

[77] MussaratYasmin, Muhammad Sharif, and SajjadMohsin. "Powerful descriptor for image retrieval based on Angle Edge and Histograms." Journal of applied research and technology 11, no. 5 (2013): 727-732.

[78] Ogle, V.E., and M., Stonebraker. "Chabot: Retrieval from a relational database of images." IEEE Comput. 28, no. 9 (1995): 40-48.

[79] Valova, I., and B. Rachev. "Retrieval by Color Features in Image Databases." ADBIS, 2004.

[80] Iqbal, Q., and J.k. Aggarwal. "Combinig Structure, Color and texture forIimage Retrieval: A performance Evaluation." International Conference on Pattern Recognition. Canada, 2002. 438-443.

[81] Pass, G., and R. Zabith. "Comparing Images using Joint Histogram." Multimedia Systems 7 (1999): 234-240.

[82] Suhasini, P.S., Dr.K.Sri Rama Krishna, and Dr. I.V. Murali Krishna. "CBIR using Color Histogram Processing." Journal of Theoretical and Applied Information Technology 6, no. 1 (2005): 116-122.

[83] Konstantinidis, K., A. Gasteratos, and I. Andreadis. "Image Retrieval based on Fuzzy Color Histogram Processing." Optics Communication, 2005: 375-386.

[84] Liu, Guang-Hai, and Jing-Yu Yang. "Contentbased image retrieval using color difference histogram." Pattern Recognition 46, no. 1 (January 2013): 188-198.

[85] Shaila, S.G., and A. Vadivel. "Block Encoding of Color Histogram for CBIR Applications." 2nd International Conference on Communication, Computing & Security. 2012. 526-533.

[86] Pass, C., and R. Zabith. "Histogram refinement for Content-based Image Retrieval." IEEE Workshop on Applications of Computer Vision. 1996. 96-102.

[87] Haung, J.et al. "Image Indexing using Color Correlogram." IEEE International Conference on Computer Vision and Pattern Recognition. 1997. 762-768.

[88] Gevers, Theo, and Arnold W.M. Smeulders. "Content-based image retrieval by viewpointinvariant color indexing." Image and Vision Computing 17, no. 7 (May 1999): 475-488.

[89] Salvador, Elena, Andrea Cavallaro, and TouradjEbrahimi. "Cast shadow segmentation using invariant color features." Computer Vision and Image Understanding 95, no. 2 (August 2004): 238-259.

[90] Cavallaro, A., T. Ebrahimi, and Elena Salvador. "Shadow identification and classification using invariant color models." IEEE International Conference on Acoustics, Speech, and Signal Processing Proceedings. (ICASSP '01). 2001. 1545- 1548.

[91] Tsai, V.J.D. "A comparative study on shadow compensation of color aerial images in invariant color models." IEEE Transactions on Geoscience and Remote Sensing 44, no. 6 (2006): 1661- 1671.

[92] Howe, N.R., and D.P. Huttenlocher. "Integrating Color, Texture and Geometry fot Image Retrieval." Proceedings IEEE Conference on Computer Vision and Pattern Recognition. 2000. 61-64.

[93] Michele, S. "Content-Based Image RetrievalLiterature Survey." Multi-Dimensional Digital signal Processing, March 18, 2008.

[94] Lazebnik, S. "A sparse texture representation using local affine regions." Pattern Analysis and Machine Intelligence 27, no. 8 (2005): 12651278.

[95] Gevers, T. " Robust histogram construction from color invariants." Proceedings. Eighth IEEE International Conference on Computer Vision, 2001. ICCV 2001. IEEE, 2001. 615620.

[96] Chatzichristofis, S. A., and Y. S. Boutalis. " Fcth: Fuzzy color and texture histogram-a low level feature for accurate image retrieval." Ninth International Workshop on Image Analysis for Multimedia Interactive Services, 2008. WIAMIS'08. IEEE, 2008. 191-196.

[97] Han, J, and K. K. Ma. "Fuzzy color histogram and its use in color image retrieval." IEEE Transactions on Image Processing 11, no. 8 (2002): 944-952.

[98] Shim, S. O., and T. S. Choi. " Image indexing by modified color cooccurrence matrix." IEEE International Conference on Acoustics, Speech, and Signal Processing (ICASSP'03). IEEE, 2003. 577.

[99] Zhao, Q., and H. Tao. "Object tracking using color correlogram." 2nd Joint IEEE International Workshop on Visual Surveillance and Performance Evaluation of Tracking and Surveillance, 2005. IEEE, 2005. 263-270.

[100] Chan, Y. K., and C. C. Chang. "A color image retrieval method based on color moment and color variance of adjacent pixels." International journal of pattern recognition and artificial intelligence 16, no. 1 (2002): 113-125.

[101] Murching, A. M., A. Tabatabai, and T. Naveen. "Histogram-based segmentation of images and video via color moments." U.S. Patent No. 6,381,363, U.S. Patent and Trademark Office, Washington, DC, 2002.

[102] Wang, T., S. M. Hu, and J. G. Sun. "Image Retrieval Based on Color-Spatial Feature [J]." Journal of Software 10 (2002): 16.

[103] Lu, Tzu-Chuen, and Chin-Chen Chang. "Color image retrieval technique based on color features and image bitmap." Information Procesing and Management 43, no. 2 (March 2007): 461-472.

[104] Zhang, Dengsheng, and Guojun Lu. "ContentBased Shape Retrieval Using Different Shape Descriptors: A comparative Study." Gippsland School of Computing and Information Technology Monash University, Churchill, Victoria 3842, 2001.

[105] MussaratYasmin, Muhammad Sharif, and SajjadMohsin. "Image Retrieval Techniques Using Shapes of Object: A Survey." Sci. Int (Lahore) 25, no. 4 (2013): 723-729.

[106] MehwishREHMANM. SHARIF, and M. RAZA. "Shape Features Extraction Method for Content based Image Retrieval." Sindh University Research Journal-SURJ (Science Series) 48, no. 1 (2016).

[107] Latecki, L.J., and R. Lakamper. "Shape Similarity measure based on correspondence of visual parts." IEEE Trans. Pattern Analysis and Machine Intelligence 22, no. 10 (2000): 11851190.

[108] Belongie, S., J. Malik, and J. Puzicha. "Shape matching and object recognition using shape contexts." IEEE Trans. Pattern Analysis and Machine Intelligence 24, no. 4 (2002): 509-522.

[109] Petrakis, E.G.M., A. Diplaros, and E. Milios. "Matching and retrieval of distorted and occluded shapes using dynamic programming." IEEE Trans. Pattern Analysis and Machine Intelligence 24, no. 4 (2002): 509=522.

[110] Chu, C.N., S.Y. Kim, J.M. Lee, and B.H. Kim. "Feed-Rate Optimization of Ball End Milling Considering Local Shape Features." CIRP Annals - Manufacturing Technology 46, no. 1 (1997): 433-436.

[111] Brandt, S. "Statistical shape features in contentbased image retrieval." Pattern Recognition Proceedings. 15th International Conference on Pattern Recognition. 2000. 1062- 1065.

[112] Weickert, J., S. Ishikawa, and A. Imiya. "Linear Sacle Space Has Been Proposed in Japan." Journal of Math,Imaging and vision 10 (1999): 237-252.

[113] Shih, Timothy K., and Lawrence Y. Deng. "Content-based Image Retrieval with Intensive Signature via Affine Invariant Transformation." Proceedings in International Symposium on Multimedia Software Engineering. 2000. 393400.

[114] Gevers, T., and A.W.M. Smeulders. "Contentbased image retrieval by View-Point-Invariant Image Indexing." Image and Vision Computing 17, no. 7 (1999).

[115] Gevers, T., and A.W.M. Smeulders. "PictoSeek: Combining Color and Shape Invariant Features for the Image Retrieval." IEEE Trans. Image Processing 9, no. 1 (2000): 102-119.

[116] Lowe, D.G. "Object recognition from local scale-invariant features." Proceedings of the Seventh IEEE International Conference on Computer Vision. 1999. 1150- 1157.

[117] Esperanca, C., and H. Samet. "A Differential Code for Shape Representation in Image Database Applications." Proceedings International Conference Image Processing. 1997.

[118] Mehtre, Babu M., and Mohan S .Kankanhall. "Shape measures for CBIR: A comparison." Information Processing & Management (Elsevier Science) 33, no. 3 (1997): 319-337.

[119] Bimbo, Del, A., and P. Pala. "Visual Image Retrieval by Elastic Matching of User Sketches." IEEE Trans. Pattern Analysis and Machine Intelligence 19, no. 2 (1997): 121-132.

[120] Pala, P., and S. ,Santini. "Image retrieval by shape and texture." Pattern Recognition 32, no. 3 (1999): 517-527.

[121] Unser, M.,,Aldroubi A., and M. Eden. " Fast Bspline transforms for continuous image representation and interpolation." IEEE Transactions on Pattern Analysis and Machine Intelligence 13, no. 3 (1991): 277-285.

[122] Haralick, R. "Statistical and structural approaches to texture." IEEE Proceedings 67, no. 5 (1979): 786-804.

[123] Li, J., A. Najmi, and R.M. Gray. "Image Classification by a two-dimensional hidden markov model." IEEE Trans.Signal Processing 48, no. 2 (2000): 527-533.

[124] Davis, L., S. Johns, and J. Aggarwal. "Texture Analysis using generalized cooccurence matrices." IEEE Trans. Pattern Analysis and Machine Intelligence PAMI-1, no. 3 (July 1979): 251-259.

[125] Unser, M., and M. Eden. "Multiresolution feature extraction and selection for feature segmentation." IEEE Trans. Pattern Analysis and Machine Intelligence 11, no. 7 (July 1989): 173-188.

[126] Maragos, P. "Pattern Spectrum and Multi-Scale shape representation." IEEE Trans. Pattern Analysis and Machine Learning 11, no. 7 (1989): 701-716.

[127] Wang, L., and D.C. He. "Texture Classification using texture spectrum." Pattern Recognition 23, no. 8 (1990): 905-910.

[128] MussaratYasmin, SajjadMohsin, and Muhammad Sharif. "Intelligent image retrieval techniques: a survey." Journal of applied research and technology 12, no. 1 (2014): 87103.

[129] Manjunath, B., and W. Ma. "Texture features for browsing and retrieval of image data." IEEE Trans. Pattern Analysis and Machine Intelligence 18, no. 8 (1996): 837-842.

[130] Unser, M. "Texture Classification and Segmentation using Wavelet Frames." IEEE Trans. Image Process. 4, no. 11 (1995): 15491560.

[131] Oraintara, S., and T. T. Nguyen. "Using Phase and Magnitude Information of the Complex directional Filter Bank for Texture Image Retrieval." Proceedings IEEE Internatinal Conference on Image Processing (October 2007): 61-64.

[132] Sastry, Challa S., M. Ravindranath, Arun K. Pujari, and B.L. Deekshatulu. "A modified Gabor function for CBIR." Pattern Recognition Letters 28, no. 2 (2007): 293-300.

[133] Kutter, M., and S. Winkler. "A vision-based masking model for spread-spectrum image watermarking." Image Processing 11, no. 1 (2002): 16- 25.

[134] Haralick, R., K. Shanmugam, and L. Dinstein. "Texture features for image classification." IEEE Trans. Systems, Man, and Cybernetics 3 (1973): 610-621.

[135] Daubechies, L. "The wavelet transform, time- frequency localization and signal analysis." IEEE Trans. Information Theory 36, no. 5 (1990): 961-1005.

[136] Chang, T., and C.J. Kuo. "Texture analysis and classification with tree-structured wavelet transform." IEEE Trans. Pattern Analysis and Machine Intelligence 2, no. 4 (1993): 429-441.

[137] Bovik, A.C., M. Clark, and W.S. Geisler. "Multichannels texture analysis using localized spatial filters." IEEE Trans. Pattern Analysis and Machine Learning 12, no. 1 (January 1990): 55-73.

[138] Kandaswamy, Umasankar, Stephanie A.Schuckers, and Donald Adjeroh. "Comparison of Texture Analysis Schemes under Non-ideal Conditions." IEEE Trans. on Image Processing 20, no. 8 (August 2011): 2260-2276.

[139] Chantler, M.J. "The effect of variation in illuminant direction on texture classification." 1994.

[140] Chantler, Mike, GedMcGunnigle, Andreas Penirschke, and Maria Petrou. "Estimating Lighting Direction and Classifying Textures." In Proceedings BMVC. 2002. 737-746.

[141] Lazebnik, S. "A sparse texture representation using local affine regions." Pattern Analysis and Machine Intelligence 27, no. 8 (2005): 12651278.

[142] Sato, Jun, and Roberto Cipolla. "Extracting the affine transformation from texture moments." Computer Vision 801 (1994): 165-172.

[143] Interrante, Victoria, and Sunghee Kim. "Investigating the effect of texture orientation on the perception of 3D shape." Proceedings in SPIE Human Vision and Electronic Imaging VI. 2001. 330.

[144] Chantler, M.J., M. Petrou, A. Penirsche, M. Schmidt, and G. McGunnigle. "Classifying surface texture while simultaneously estimating illumination direction." International Journal of Computer Vision 62, no. 1 (2005): 83-96.

[145] Ojala, T., M. Pietikäinen, and T. Mäenpää. "Multiresolution gray-scale and rotation invariant texture classification with local binary patterns." IEEE Trans. Pattern and Machine Learning 24, no. 7 (July 2002): 971-987.

[146] Leung, T., and J. Malik. "Representing and recognizing the visual appearance of materials using three-dimensional textons." International Journal of Computer Vision 43, no. 1 (2001): 29-44.

[147] Hahjidemetriou, E., M. Grossberg, and S. Nayar. "Multiresolution histograms and their use for recognition." IEEE Trans. Pattern Analysis and Machine Intelligence 26, no. 7 (2004): 831-847.

[148] Malik, J., and R. Rosenholtz. "Computing Local Surface Orientation and Shape from Texture for Curved Surfaces." International Journal of Computer Vision 23, no. 2 (June 1997): 149168.

[149] Texture features for browsing and retrieval of image data. Manjunath, B. and Ma, W. 8, 1996, IEEE Trans. Pattern Analysis and Machine Intelligence, Vol. 18, pp. 837-842.

[150] Ponce, J., S. Lazebnik, and C. Schmid. "A maximum entropy framework for part-based texture and object recognition." Computer Vision, ICCV Tenth IEEE International Conference on Computer Vision. 2005. 832838.

[151] Schaffalitzky, F., and A. Zisserman. "ViewPoint invariant texture matching and wide baseline stereo." Proceedings IEEE ICCV, 2001.

[152] Ponce, J., Lazebnik, S. andSchmid,"Affineinvariant local descriptors and neighborhood statistics for texture recognition." Computer Vision Proceedings. 9th IEEE International Conference. Oct. 2003. 649- 655.

[153] Porter, R., and N. Canagarajah. "Robust rotation-invariant texture classification: Wavelet, Gabor filter and GMRF based scehemes." IEEE Process Vision Image Signal Process 144, no. 180 (June 1997).

[154] Laine, A., and J. Fan. "Texture Classification by Wavelet Packet Signature." IEEE Trans. Pattern Analysis and Machine Intelligence 15, no. 11 (1993): 1186-1191.

[155] Daubechies, J. Ten Lectures on Wavlets. Philadelphia: SIAM, 1992.

[156] Chui, C.K., L. Montefusco, and L. Puccio. Wavelets: Theory,Algorithmsand applications. Academic Press, 1994.

[157] Randen, T., and J.H. Husoy. "Filtering for Texture Classification: A Comparative Study." IEEE Trans. Pattern Analysis and Machine Intelligence. 1999. 291-310.

[158] Coifman, R.R., and M.V. Wickerhauser. "Bestadapted wavelet packet bases." Yale University preprint, 1990.

[159] Coifman, Ronald R., Yves. Meyer, Steven. Quake, and M. Victor. Wickerhauser. "Signal processing and compression with wavelet packets." Wavelets and Their Applications. 1994. 363-379.

[160] Daubechies, I. "Orthonormal bases of compactly supported wavelets." Communication Pure Applied Mathematics 41 (1988): 909-996.

[161] Ramchandran, K., and M. Vetterli. "Best wavelet packet bases in a rate-distortion sense." Image Processing, IEEE Transactions 2, no. 2 (1993): 160- 175.

[162] Banerjee, Minakshi, and Malay K. Kundu. "Content-Based Image Retrieval Using Wavelet Packets and Fuzzy Spatial Relation." ICVGIP (Springer-Verlag Berlin Heidelberg), 2006: 861-871.

[163] Mallat, S. "A theory for multi-resolution signal decomposition." IEEE Trans. on Pattern Analysis and Machine Intelligence 11, no. 7 (1989): 674-693.

[164] M.Acharya, and M.K.Kundu. "An Adaptive approach to unsupervised texture segmentation using m-band wavelet transform." Signal Processing 81 (2001): 1337-1356.

[165] M.A.Hoang, J.M. Geusebroek, and A.W.M.Smeulders. "Color texture measurement and segmentation." Signal Processing 85 (2005): 265-275.

[166] Rao, R.M., and A.S. Bopardikar. Wavelet Transforms Introduction to Theory and Applications. Singapore: Pearson Education, 2002.

[167] Cullen, J.F., J.J Hull, and P.E. Hart. "Document Iamge Database Retrieval and Browsing Using Texture Analysis." Proceedings Fourth International Conference Document Analysis and Recognition. 1997. 718-721.

[168] Ma, W.Y., and B.S. Manjunath. "Edge Flow: A Framework of Boundary Detection and Image Segmentation." Proceedings Computer Vision and Pattern Recognition. 1997. 744-749.

[169] Li, C.S, and V. Castelli. "Deriving Texture Feature Set for Content-Based Image Retrieval of Satellite Image Database." Proceedings International Conference on Image Processing. 1997.

[170] Wang, Xing-Yuan, Zhi-Feng Chen, and JiaoJiao Yun. "An effective method for color image retrieval based on texture." (Computer Standards & Interfaces) 34, no. 1 (2012): 31-35.

[171] Dana, Kristin J., and Shree K. Nayar. "Correlation Model for 3D Texture." International Conference on Computer Vision. 1999. 1061-1067.

[172] MussaratYasmin, Muhammad Sharif, and SajjadMohsin. "Use of Low Level Features for Content Based Image Retrieval: Survey." Research Journal of Recent Sciences

[173] MehwishRehman, Muhammad Iqbal, Muhammad Sharif, and MudassarRaza. "Content based image retrieval: survey." World Applied Sciences Journal 19, no. 3 (2012): 404412.