Content area
Full Text
Abstract
Recently, image rotation has cecome very important preprocessing step for image processing and pattern recognition. It involve multiplication of the rotation and is thus very show Herein. We propose a high-speed image processing algorithm that eliminates both this delay and problems with image matching. We compare our algorithm with conventional approache and describe some basic requirements for noise removal algorithms in projector. The proposed algorithm can perform faster and more accurate image mapping than would be possible with a CPU through high speed image processing and feature extraction than .
Key Words: Image Processing, Transform Matrix, High Speed Rotation, Medical Image Processing
(ProQuest: ... denotes formulae omitted.)
1.Introduction
Today, computers are able to process-а huge amount of data, and - complex mathematical algorithms are used in information systems and datacenters in almost all areas of science and technology. High computer performance is required, and it is clear that for such calculations the performance of common personal computers is insufficient [1,2]. Moreover, users in the science industry who do not work with complex computer calculations are not interested in optimization and therefore tend to use more resources than that are necessary. In addition to image processing, it is also effective for other computationally intensive tasks. We explain why CUDA is ideal for image processing, and how efficiently it can extract features from iris images. This method eliminates the need to calculate a large number of mathematical and trigonometric functions [3, 4, 5], replacing them with a linear transformation to reduce and simplify the computation required and enable high-speed image rotation. The CUDA Processor class also implements the cudaCall abstract method, which provides an entry point for calling CUDA kernel functions. This structure allows the integration of CUDA image processing algorithms with a minimum of overhead and maximum code reuse [5]; therefore, it is used for our high-speed algorithm. In this paper, we investigate two types of method, namely matrix and shearing methods. We use a matrix method, involving trigonometric functions, and a shearing method, a simple rotation method involving a linear transformation. The shearing method, previously introduced in order to speed up the matrix method, calculates the trigonometric function values for the corresponding slopes in advance and stores them in a table. We compare...