基于多维矢量矩阵离散余弦变换的熵编码

针对传统JPEG中游程-霍夫曼熵编码方法在彩色图像的多维矢量矩阵离散余弦变换体系中,不能充分有效压缩数据的问题,提出了一种根据非零交流系数幅值尺寸和该系数前零游程长度尺寸的联合分布进行游程编码,再进行霍夫曼编码的新方法,并重新统计设计了霍夫曼码表。实验结果表明:该方法的压缩性能较改进前熵编码,在PSNR相同的情况下,比特率至少降低了9%,远优于JPEG方法,在更高维的视频图像变换压缩中有着良好的应用前景。

Entropy code based on multidimensional vector matrix DCT

Run length-Huffman coding(RL-VLC) in JPEG is normally used as entropy coding method for multi-dimensional vector matrix discrete cosine transform(MD-VMDCT).However it can not effectively compress image since there are lots of long run length of quantized coefficients after three-dimensional scan,here an improved RL-VLC coding method was proposed.According to the Huffman code table which is designed by the probability distributions of level and run level combination,this algorithm achieved entropy coding for color image.Experimental results show that the compression efficiency of proposed algorithm is better than traditional Huffman coding,in the case of the same PSNR,the proposed algorithm saves bit rate at least 9%,much better than JPEG,and there are good prospects for the application of higher-dimensional video frames compression.