结构信息最优的静止图像压缩算法研究
doi: 10.3724/SP.J.1146.2009.00983
Still Image Compression Algorithm Based on Structural Information Optimization
-
摘要: JPEG2000是基于小波变换的新一代静止图像压缩标准，与以往的压缩标准相比，其具有很多优点。但是JPEG2000以MSE作为图像失真评价标准，而MSE不能很好的符合人眼主观评分，进而很大地影响了JPEG2000的压缩性能。该文在JPEG2000标准框架下，提出了以结构相似度作为失真评价标准的静止图像压缩算法(SJPEG2000)。该算法以系数对图像结构信息贡献量的大小作为准则来截取码流，使压缩后的图像尽量保存原图像的结构信息。实验结果表明，该算法压缩得到的图像很好地保留了图像结构信息，压缩图像的主观质量得到提高，结构相似度值较原JPEG2000也有一定提高。Abstract: JPEG2000 is a new still image compression standard based on wavelet transform. Compared with previous compression standards, it has a lot of advantages. However, in JPEG2000 the image distortion evaluation criteria is MSE, and MSE can not correlate very well with subjective ratings. So compression performance of JPEG2000 is affected greatly. In this paper, a still image compression algorithm SJPEG2000 that uses SSIM (Structural SIMilarity) as the distortion evaluation criteria is proposed under the framework of JPEG2000 standard. In order to let the compressed images retain more structural information, the algorithm intercept stream according to the contribution of the image structural information. Experimental results show that the images compressed by this algorithm retain much more structural information than original JPEG2000 and the corresponding SSIM value is also improved.
-
计量
- 文章访问数: 3399
- HTML全文浏览量: 91
- PDF下载量: 955
- 被引次数: 0
下载:
