一种新的基于DSP应用的改进SPECK算法

在基于小波变换的图像压缩编码算法中，Pearlman的利用子带内小波系数的聚类特性的Set Partition Embedded block(SPECK)算法与SPIHT算法相比，具有更低复杂度、更快的编解码速度和相近的性能。文中提出了一种基于DSP平台的SPECK编码器在实时环境下的应用方案。通过使用提升结构的整数小波变换并基于DSP平台上对变换流程进行优化，以提高小波变换的速度。引进误差比特数(Number of Error Bits)概念，并定义绝对零系数(Absolute Zero-Coefficient)对原有的SPECK算法进行改进，在不影响压缩性能的情况下，显著地减少了原算法对内存的需求，并提高了执行速度。实验结果证明，改进后的算法适应了大多数的实时系统的要求，是一个具有实用价值的DSP解决方案。

Real-time Implementation of a New SPECK Image Codec Using DSP Chip

Among all algorithm based on wavelet transform, Pearman's set partition embedded block(SPECK) algorithm is well known for its simpler structure and faster coding speed than SPIHT with similar efficiency to SPIHT. This paper deal with the real time implement of SPECK algorithm using DSP chip. In order to facilitate the implementation and improve the codec's performance, some relative techniques will be discussed, such as the optimization of program structure of wavelet decomposition to increase speed of wavelet filter. With the concept of Number of Error Bits introduced and a new concept of Absolute Zero-Coefficient presented, a new way different with the original algorithm to control coding will be provided and thus the memory cost and the coding time is significantly reduced without any performance lost . Our experimental results show that the implementation meets common requirement of real-time video coding and is proven to be a practical and efficient DSP solution .