Parallel Optimization Technology Of Satellite Image Decompression Based On Multi-core Processor

This thesis relates to a parallel optimization technology of real-time decompression software for satellite images running on a universal multi-core processor.The decompression software receives and downloads the compressed code stream data,and completes the image reconstruction tasks such as code stream frame analysis,decoding,storage and display in real time.Among them,the decoding link involves data processing that is computationally intensive and frequently accessed.The traditional serial implementation takes a long time and becomes a bottleneck restricting the real-time performance of decompression and reconstruction.This thesis makes full use of the advantages of multi-core processor parallel processing,and optimizes the decompression software from two aspects: code stream frame parallel and data block parallel.The specific contents are as follows:To solve the problem of low throughput of decompressed data caused by the time-consuming decoding process,a decoding optimization method based on code stream frame parallel is proposed.By modifying the original pipeline computing structure,multiple blocking threads are created to decode the multi-frame stream with continuous sequential input in parallel.Methods such as memory access optimization,memory management optimization,and compiler optimization are used to make full use of the performance of the multi-core CPU,which greatly reduces the average decoding time.Experimental results show that when Intel i9 processs 10 cores in parallel,the average decompression frame rate increases from 0.11 frames/sec to 1.71 frames/sec.To solve the problem of large decompression delay caused by the long time-consuming RS error correction and JPEG-LS decoding,a decoding optimization method based on data block parallel is proposed.The data parallel between RS error correction and JPEG-LS decoding is implemented by Open MP programming model.Computing tasks are evenly distributed to multiple different threads with a smaller parallel granularity and dynamic scheduling mode to achieve load balancing.Experimental results show that when Intel i9 processs 10 cores in parallel,the decompression time is reduced from 4.98 seconds/frame to 2.02 seconds/frame.To solve the problem of long decompression time of code stream frame parallel and low decompression frame rate of data block parallel,a two-level parallel decoding optimization method combining stream frame parallel and data block parallel is proposed.Through the Qt thread library and the Open MP programming model,the multi-frame code stream and the single-frame code stream are decoded for data parallelism.On Intel i9processor(16 cores in total),it balances the decompression time and the decompression frame rate.Experimental results show that the average frame decompression frame rate is increased from 0.11 frames/sec to 1.59 frames/sec,and the decompression time is reduced from 4.98sec/frame to 2.59sec/frame,which meets the requirements of decompression tasks.