Research And Implementation Of Parallel Algorithms. Multi-texture Image Segmentation

Texture is one of the most important feature in describing the image structure and content, it is an important research content of computer vision and pattern recognition.Texture segmentation is not only the basis of texture analyzing,but also one of the most important links of image segmentation, image restoration, image enhancement, image registration, image retrieval, image classification and so on.Human's visual system can be considered as a multi-channel filter model, and the 2-D Gabor filter formed by Gabor function is such a filter model. So the method of texture segmentation based on Gabor filter is becoming the focus of texture segmentation research. As the texture in our lives is becoming more and more complex, it is difficult to complete the process of texture segmentation of complex image in a limited time by single-core processor, the effective way is the parallel processing.This paper improved Wu Gaohong and Zhang Liujin' method of constructing the best Gabor filter. When finding the most different area between two kinds of texture's power spectrum, we introduced a matrix, it is feasible in not only double texture, but also multiple texture. We implement a multi-threads parallel segmentation algorithm by adopting the technology of OpenMP parallel programming interface. And on the basis of above discussion, we discussed another parallel method on cluster using multi- processes and multi-threads by the combination of OpenMP and MPI. It reduced computing time and improved efficiency.Some experimental results show that the parallel methods can make full use of the computation resources, produce the texture image segmentation quickly and save a lot of computing time.