基于分级子空间回归的压缩人脸图像复原-《计算机技术与发展》

文章信息/Info

Title:: Compressed Face Image Restoration Based on Hierarchical Subspace Regression

Author(s):: LIU Xin-yu; GAN Zong-liang; LIU Feng; School of Telecommunications &Information Engineering,Nanjing University of Posts and Telecommunications,Nanjing 210003,China

Keywords:: compressed face images; subspace regression; edge-orientation; shallow subspace; deep subspace

摘要:: 人脸图像具有自然图像不具备的对称特征和几何结构相似性。由于人脸图像往往具备非常复杂的角点和纹理特征,因此很难找到一种全局模型将压缩图像映射到原始未压缩图像。针对此问题,提出一种新颖的基于分级子空间回归的压缩人脸复原算法,该算法包括训练和复原两个部分。在训练部分,利用压缩人脸图像的边缘方向分布规律,将压缩- 未压缩图像块对划分到多个浅层子空间中。然后对每个基于边缘方向分类的浅层子空间,利用 K-means 聚类算法得到它的深层子空间,并在每个深层子空间中训练得到相应的线性映射。在复原阶段,对每个输入的压缩图像块分析得到它的边缘方向,从而选择合适的线性映射,得到复原后的输出图像块。实验结果表明,该算法在 PSNR 和 SSIM 上均优于现有的常用复原算法,并且能够有效地去除压缩失真和锯齿效应,提高视觉效果。

Abstract:: Face images have the character of symmetry and similar geometry structures that are not available in natural images. Since face images often have very complex corner points and texture feature,it is difficult to find a global model to map compressed images to raw uncompressed images. Aiming at this problem,we propose a novel compressed face restoration algorithm based on hierarchical subspace regression,which includes two parts:training and restoration. In the training,the rule of the face edge-orientation distribution is used to classify the compressed-uncompressed patch pairs into shallow subspaces. Then,the K-means clustering is used to cluster the deep subspaces of each shallow subspace,and corresponding linear mapping training is performed for each deep subspace. In the restoration,an appropriate linear mapping selected based on the edge orientation of compressed input image patch is applied to generate the restored output image patch. The experiment shows that the PSNR and SSIM are better than the existing popular algorithm,and it can effectively remove the blocking artifact and zigzag effect,so as to improve the visual effect.