基于自适应高维度非局部总变分的压缩图像去块效应算法

Image Deblocking Algorithm Using Adaptive High-dimensional Nonlocal Total Variational for Compressed Image

经JPEG压缩后的图像存在着块效应,尤其在低码率段块效应更加明显,严重影响了图像质量和视觉效果。通过引入压缩噪声的概念,利用压缩图像及压缩比特流中获得的信息,对压缩过程进行建模。然后在最大后验框架下,联合自适应高维度非局部总变分先验及窄量化约束先验构建代价函数,提出了一种能有效去除压缩图像块效应的后处理算法。实验结果表明,相比于一些主流的去压缩效应的算法,本文算法的处理结果有更高的客观参数及更好的主观视觉效果。

JPEG-compressed images have block artifacts,especially in the low bit rate.By introducing the concept of compressed noise,the compression process is modelled by the information obtained from the compressed image and the compressed bit stream.The algorithm proposed is a post-processing method which can effectively remove the effect of compressed image block under the maximum posteriori framework.It combines an adaptive high-dimensional nonlocal total variational and a narrow quantization constraint to solve the problem.The experimental results show that the method has better objective experimental results and subjective visual effects than other reducing blocking artifacts algorithms.