Parallelization Research Of Face Recognition And Texture Synthesis Algorithm Base On MATLAB Multi-core Clusters

In order to take reasonable advantage of multi-core clusters and improve the available efficiency of CPU, parallel optimization was studied which according to the algorithm of PCA-based face recognition and Image Quilting texture synthesis based on MATLAB distributed parallel computing cluster. Deeply analyzing the cluster architecture, the hardware and software system structure, algorithm programming model and so on, and combined with structure features of serial program complete the parallel programming process, which can play a well potential capacity of multi-core and promote the performance of multi-core application.The research focus is as follows:For one thing, according to the algorithm of PCA-based face recognition, parallel programming model by building blocks with multithreading was studied to improve the performance of the program. According to the integral structure of PCA-based face recognition algorithm, a functional module named train() was designed for the training of recognizing generated samples in the environment of MATLAB cluster. The parallelization of this algorithm was realized by task partition. The experimental results indicate that the stable recognition rate of94.167%and the approximately linear speedup ratios verify the correctness and the high efficiency of the algorithm parallelization, and the development of parallel application was condensed further. Parallel method caused little influence of the algorithm through the MATLAB distributed parallel computing cluster, and better solved intensive computing optimization problems of large amount of data in multi-core computer or multi-core clusters. For another thing, according to Image Quilting for texture synthesis algorithm characteristic that the new and old texture pieces spliced, a parallel programming model that creating multi-task realized the texture pieces to cut and splice was studied to Image Quilting for texture synthesis algorithm. A functional module named iq() was designed for the calculation of suture path in the overlapping area, subtasks assigned to each Worker executed parallel computing by using of the scheduler of MATLAB multi-core clusters. The experimental results indicated that this algorithm can achieve a better speed-up ratio and enhance the available efficiency of multi-core CPU and promote the utilization rate of multi-core computer resources effectively.The characteristics of the research was to design the MATLAB parallel applications through multi-thread creating, dynamic configuration task manager, thread number setting and thread scheduling strategy in order to improve the algorithm performance. In addition, the attention of algorithm’s improvement and development should be paid to data sharing, module partition, task parallelism and distribution and load balancing between processor core numbers, which to ensure each nucleus of the multi-core processors in a high load condition and accomplish parallel computing task accurately.It effectively modified the bottleneck of serial program in the algorithm of PCA-based face recognition and Image Quilting texture synthesis through theverification of MATLAB multi-core cluster platform, which have improved the working efficiency of CPU and made full use of multiple nuclear resources. At the same time, it also demonstrated the advantages that algorithm application of graphic and image combining with multi-core parallel computing technology.