Parallel Processing Of Edge Detection Algorithms Of Images Corrupted By Noises Based On HyperX Multi-core Processor

With the continual development of computer science and technology, image processing has made tremendous progress in terms of processing precision and accuracy. Meanwhile, big data information impose great pressure to the the traditional single-core processor. Thus, under the precondition of ensuring the precision and accuaracy, improving the speed of image processing has become the most significant research interest.In image processing, edge information of the image depict the basic characteristics, which is the most important information we needed. The traditional edge detection algorithms,such as Differential operator method, Sobel operator, Robert operator, Prewitt operator, Canny operator and so on, can extract the image edge information easily and quickly. However, these state of art algorithms have common demerit that is weak resistance of noise. Aiming at the image contaminated by the noise severely, these traditional edge detection algorithms can be hardly employed due to the false detection and missed information of the image edge. This paper proposes a novel algorithm which combines the Bilateral Filtering and Canny Algorithms. The new algorithm is implemented through the software platform named Matlab 2014 a can not only make sure the accuracy in image edge detection but also reduce the influence of different noises. Furthermore, we port it into the multicore platform named HyperX. This novel algorithm takes advantage of the superiorities of preserving edge information and reducing noise of the Bilateral Filtering algorithm, which also utilize the accuarcy of the Canny edge detection algorithm. Experimental results demonstrate that detection performance has improved dramaticlly. The detailed steps are as follows:We utilize the Bilateral Filtering to reserve the edge information and reduce the noise information for the target image firstly, and then for the processed image, the Canny edge detection algorithm is adopped to detect the edge information.Experimental results implies that the algorithm we proposed can effectively improve the accuracy of edge information extraction for the images polluted by the noise seriously. Due to that the target image contains plentiful data information, it is necessary to improve the speed of edge detection. Therefore, we port the proposed algorithm into the HyperX high-performance parallel platform. The design process is described in detail:Firstly, we accomplish the split of the algorithm through using the FAST design mechanism provided by the HyperX platform;Secondly, we design the mechanism of the routing communication and the synchronization mechanism to prevent data channel and instructions execution from blocking;At last, we compile the configutation profiles to make sure the proper storage of data, creation of library files, design of the PE information and other accurate configuration profiles.Experimental results indicate that the scheme has the reliable stability. The novel algorithm can be implemented effciently and quickly in parallel model by the premise of the precision and the accuracy. In addition, the design stucture can be used with other images processing algorithms. In summary, the design structure has generality.