带等式约束的总变分JPEG解码算法

A Total Variation Based JPEG Decompression Algorithm With Equality Constraint

针对低比特JPEG图像因量化取整过程中产生的量化噪声问题,提出一种带等式约束的总变分(TV)模型。首先根据JPEG算法的量化取整得到一个量化约束和等式约束;然后选择L1范数作为数据保真项;最后利用Split Bregman算法求解TV模型得到后处理JPEG解码图像。实验结果表明,本文方法得到的解码图像优于之前的总变分方法。

Focused on the problem that quantization noise in the quantization process of low-bit JPEG images, a total variation （TV）model with equality constraint was proposed. Firstly, a quantization constraint was gotten by quantization and round of JPEG algorithm ; Secondly, selected L1 norm as data fidelity term ; Finally, the Split Bregman algorithm was exploited to solve the proposed TV model and so that we obtain a post-processing JPEG images. The experimental results demonstrate that the decompression images with proposed method yield better than state-of- the-art total method.