%0 Journal Article
%T 基于PMP和SSR的图像去块算法
Image Deblocking Algorithm Based on PMP and SSR
%A 李致运
%A 于之雅
%A 马梓锐
%A 姜涵月
%A 王名洋
%A 李喆
%J Advances in Applied Mathematics
%P 83-99
%@ 2324-8009
%D 2025
%I Hans Publishing
%R 10.12677/aam.2025.144142
%X 块离散余弦变换(BDCT)编码在图像和视频压缩领域应用广泛。然而，在低码率编码条件下，压缩图像的块边缘常会出现明显的块效应，这极大地影响了图像的视觉效果。为此，本文提出了一种结合局部最小像素(PMP)正则化和结构稀疏表示(SSR)的方法，旨在去除压缩图像中的块状伪影，同时保留图像的锐利边缘和细节信息。具体而言，我们利用内部结构稀疏先验来消除图像噪声，并借助外部结构稀疏先验防止图像过拟合。此外，通过实施局部最小像素正则化约束，能够有效区分块化图像和清晰图像，增强块化图像的恢复效果。在处理所提模型的非凸性问题时，我们在交替迭代法中融入了滤波技术。实验结果表明，该算法在客观性和视觉感知方面均达到了与当前几种先进去块算法相当的水平。
Block discrete cosine transform (BDCT) coding is widely used in image and video compression. However, under low bit rate coding conditions, the block edges of compressed images often show obvious block effect, which greatly affects the visual effect of the images. To this end, this paper proposes a method that combines local minimum pixel (PMP) regularisation and structured sparse representation (SSR) with the aim of removing block artefacts from compressed images while preserving the sharp edges and detail information of the images. Specifically, we remove image noise using an internal structural sparse prior and prevent image overfitting with the help of an external structural sparse prior. In addition, by implementing the local minimum pixel regularisation constraint, we are able to effectively distinguish blocked images from clear images and enhance the recovery of blocked images. When dealing with the non-convexity problem of the proposed model, we incorporate the filtering technique in the alternating iteration method. Experimental results show that the algorithm achieves a level comparable to several current state-of-the-art deblocking algorithms in terms of objectivity and visual perception.
%K 结构稀疏表示，
%K 局部最小像素，
%K 图像去块
Structured Sparse Representation
%K Local Minimum Pixel
%K Image Deblocking
%U http://www.hanspub.org/journal/PaperInformation.aspx?PaperID=110953