Research And Improvement Of CCSDS Decoding Complexity Based On SSE Technology

As the demand for better image compression algorithm rises, low delay and complexity have become important measurements for an algorithm. Pentium III and even newer processor, released by Intel?, add some new instructions: Streaming SIMD Extensions (A Single Instruction, Multiple Data Streams Extended instruction set, called SSE). The new instruction set greatly improves the performance of graphics, animation, audio and video. At the same time, eight 128-bit registers (xmm0-xmm7) are also added, which can simultaneously handle four single-precision floating-point variables thus greatly enhance the efficiency of floating-point calculations. The Consultative Committee for Space Data System (CCSDS) in November 2005 released a set of image compression algorithm standard used in the field of space, the basic unit of which includes Discrete Wavelet transform (DWT) and Bit-Plane Encoder (BPE). 9/7 wavelet is used in DWT. The algorithm is mainly used for space, for example, in the spacecraft, so it has strict requirement on real-time. Therefore, research on the complexity of CCSDS algorithm is very important. A simple, fast and small memory usage algorithm will increasingly meet the needs of the space.This paper presents the theory of CCSDS algorithm, SSE technology instruction set and the rules of calculation. Then Intel? VTune? Performance Analyzer is used to study the time complexity of the process of decoding. It is found that Inverse Discrete Wavelet transform (IDWT) including more floating-point calculations is the module that takes the most time of the entire process. As a result, according to the characteristics of SSE instruction set and IDWT, CCSDS decoding algorithm optimization solution based on the SSE instruction is proposed and realized in this paper, which reduces the time complexity of decoding. It is proven by the performance test that IDWT decoding module can save up to 86% of the time, and the entire CCSDS decoding process saves time up to 75% without changing the decoded quality of the image.