基于自适应指数哥伦布编码的图像压缩算法  

Image compression algorithm based on adaptive exp-Golomb coding

作　　者：孙健[1,2] 任国强[1] 吴钦章[1] 

机构地区：[1]中国科学院光电技术研究所,四川成都610209 [2]中国科学院大学,北京100149

出　　处：《光学精密工程》2013年第11期2973-2979,共7页Optics and Precision Engineering

基　　金：国家自然科学基金资助项目(No.11178004)

摘　　要：在徐勇等人提出的适用于硬件的低复杂度算法的基础上,提出了一种基于自适应指数哥伦布编码的图像压缩算法来进一步提高压缩性能.首先,对图像进行4级5/3小波变换;根据小波变换后的子带数据进行建模,得到最佳量化步长.然后,采用JPEG_LS算法预测量化后的IL子带,并对各个子带数据进行零游程编码.最后,对零游程编码得到的数据进行自适应指数哥伦布编码.实验表明：当比特率大于0.25 bpp时,本算法略好于徐的算法;当比特率小于0.25bpp时,本算法重构图像的峰值信噪比较徐的算法高0.2～2 dB.结果显示,本算法不仅提升了压缩性能,而且由于指数哥伦布编码级数更新模型复杂度很低,完全可以用硬件实现.According to the low-calculation image compression algorithm introduced by Xu Yong, et al. , a new image compression algorithm based on adaptive Exp-Golomb coding was proposed to im- plement the high-speed image compression. Firstly, the image with 4-level 5/3 was transformed with wavelet transform to obtain the optimal quantization step for each wavelet subband according to wave- let subband data model. Then, the JPEG LS algorithm was employed to predict the value for the LL subband and to encode for all the wavelet subband data in the Zero-Run-Length algorithm. Finally, the adaptive Exp-Golomb coding was adopted to encode the data after Zero-Run-length algorithm. Ex- periment results show that the proposed algorithm can achieve a little better performance than Xu＇s al- gorithm, when the bit rate is greater than 0.25 bpp; and the Peak Signal-to-noise Ratio （PSNR） of the proposed algorithm can increase by 0.2--2 dB when the bit rate is less than 0.25 bpp. It means that the algorithm improves the performance of compression. The Exp-Golomb coding model is ex- traordinary simple and can be implemented by the hardware completely.

关 键 词：图像压缩 小波变换 量化建模 自适应指数哥伦布编码 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]