适于硬件实现的无损图像压缩

摘要

A hardware implementation method for lossless image compression is proposed to overcome the difficulties of embedded wavelet coding methods in hardware implementation and high costs. Firstly, the algorithm divides wavelet coefficients into a low frequency block and three high frequency blocks according to sub-band properties, and then uses different methods to code respectively. In the low frequency block coding method, the Difference Pulse coding Modulation(DPCM) is firstly used to reduce coefficients' dynamic range. Then, a modified bit plane coding method is used to output the bit stream. In the high frequency block coding method, the proposed modified Set Partitioning in Hierarchical trees(MSPIHT) algorithm is used to code three high frequency blocks respectively with their thresholds. The MSPIHT optimizes the outputted bit stream by using a type of A set judge, reduces memory requirement and avoids memory dynamic management by eliminating the lists of SPIHT algorithm. Moreover, the MSPIHT avoids repeated calculation in scanning process and enhances the coding efficiency by adopting MMVS. Experiment results show that the bit-rates of all international standard testing images have reduced more than 1.4 bit/pixel and the coding speed has increased more than three times as compared with that SPHIT. It is concludes that the proposed algorithm is super in realtime performance, low memory requirement and fit for hardware implementation.%针对常见嵌入式小波编码算法硬件实现困难、成本较高等问题,提出了一种适用于硬件实现的无损图像压缩算法.该算法根据子带属性的不同将小波系数分为1个低频子块和3个高频子块,然后使用不同的方法分别进行量化编码.对于低频子块,首先使用脉冲差分编码调制(DPCM)方法压缩其数据动态,然后使用改进的比特位平面编码算法编码输出对应码流;对于各高频子块,则使用提出的改进集合树分裂(SPIHT)算法分别进行量化编码.在改进的SPIHT算法中,通过加入A类集合的分类优化了码流输出;通过消除链表,降低了内存需求并避免了内存的动态管理;通过使用集合极值矩阵,避免了扫描过程中的重复判断,提高了编码效率.实验结果表明,与传统SPIHT算法相比,本文算法可使各国际标准测试图像的编码比特率均降低0.14 bit/pixel以上,而编码速度提高3倍以上.该算法具有实时性高、内存需求低、适于硬件实现的特点.