Parallel Image Compression Based On Multi-core SMP Clustering

With the extensive use of high-definition camera,the resolution of the image taken by the camera continues to be improved.Although the image has high quality,its transmission seems more and more insufficient in the current network environment,especially distinctly under a circumstance that high real-time demand is required.Therefore,the problem how to deal with the process of the image information with larger amount,better quality of the image and larger space occupation has become a key issue and the solution has become a core technology in the existing computer and network environment,wherein the image compression is one of the solutions to deal with this problem.On the premise of ensuring the quality of the image,removing the redundancy encoding through a variety of encoding or other techniques is to acquire an effect that space occupation is reduced and the communication speed is accelerated.Some high compression ratio algorithms are on the cost of a large number of computing time which is not worthwhile undoubtedly.Therefore,shortening the time for compression has to be done when the high compression ratio is obtained.Through parallel image compression,compression speed can be improved and the compression time can be decreased.People adopt multi-core processors and clusters to improve computing speed and use a parallel processing technology to improve architecture of computer system so as to cost less time to complete computing tasks as far as possible and fully play the role of parallel action in computing.The main contents are as follows.Combined with the hierarchical structure of SMP cluster,MPI message passing programming model and OpenMP shared memory programming models are applied to a parallel design that message passing is adopted to intra-nodes communicate and shared memory is adopted to inter-nodes communicate.In an experimental environment,a Dawning Cluster TC-5000 is used as the hardware,and the parallel design of fractal image compression is carried out in a master/slave mode.In order to improve the efficiency of compression,the author respectively used pure MPI parallel,combining with the OpenMP hybrid parallel and dynamic task partition at the same time.Moreover,in the experimental part,the author first built a parallel environment and used serial,MPI parallel and fractal image compression algorithm to compress the images;then the author compared time spent by compression and analyzed improvement rules of image compression methods by parallel algorithm.First,the control behavior of the master node is improved,only the block number and the control command are distributed.The operation data is not divided by the master node.The compute node uses the data on the master node according to the instruction and number.Secondly,it puts forward the strategy of using the dynamic distribution of the task,and makes full use of the computational power of the computing node,improving the parallelism of the program and improving the robustness of the algorithm.Third,the use of packaging operations in the calculation of the node to complete the preliminary summary of the data to avoid frequent data transfer operations,reducing the communication between nodes.Fourth,the computation node uses OpenMP to map the block,transforms the gray scale code,improves the computation speed of the algorithm,increases the parallelism and reduces the communication expense.In the multi-core SMP cluster,the parallelization was carried out and the test environment was debugged.Contrast experiments show that the parallel algorithm achieves optimal block matching with less time.Which shows the important application value of parallel computing in image compression field in multi-core SMP cluster environment.