基于总变分的MPEG解码算法

针对低比特MPEG图像序列出现的压缩痕迹，提出一种后处理正则化方法进行MPEG解码。首先由自适应量化得到一个原始视频DCT系数的量化区间，重建的视频序列投影到此区间内，其次视频是连续的静止图像，得到耦合时间维度和分离时间维度的两种总变分模型，最后利用经典的原-对偶算法求解提出来的凸优化模型，得到后处理MPEG解码视频序列。实验结果表明，总变分正则化函数能够一定程度上减轻压缩痕迹，提高解码视频的质量。

Total variation based MPEG decompression algorithm

Focused on the defect that compression artifacts in low-bit MPEG images sequence, a post-processing technique of regularization was proposed to decompress MPEG objection. Firstly, the quantization interval of the DCT coefficients of video original can be gotten because of adaptive quantization , and the reconstructed video sequences projected onto the interval; Secondly, because of videos are continuous still images sequences, two total variation models that all dimensions are coupled and temporal dimension is decoupled from two spatial dimensions were presented ; Finally, a classical primal-dual algorithm was exploited to solve the convex model so as to we obtain a post-processing MPEG video sequences. The experimental results demonstrate that the total variation functions can reduce compression artifacts to some extent and improve the quality of the decoded videos.