边缘保持正则化低码率压缩图像后处理算法  

Post-processing of low rate compressed image using edge preserving regularization

作　　者：谭毅华[1] 田金文[1] 柳健[1] 

机构地区：[1]华中科技大学图像识别与人工智能研究所多光谱信息处理技术国家重点实验室,湖北武汉430074

出　　处：《光学精密工程》2007年第4期592-598,共7页Optics and Precision Engineering

基　　金：湖北省自然科学基金项目(No.2006ABA088)

摘　　要：从图像恢复的角度,提出以正则化方法完成后处理任务。分析了正则化方法的模型,并给出了边缘保持的正则化函数所应具有的特性。从复杂性、健壮性和对边缘细节粒度控制的能力三个方面选择了相应的势能函数,然后以半二次正则化将能量函数进行转换,使其快速达到最小化。最后给出了整个交替迭代后处理算法的描述。该方法对图像边缘细节具有自适应性,并能较快地取得最小值。实验结果显示,该算法能有效地提高低码率压缩图像的客观质量和视觉效果。A regularization method is proposed to implement the post-processing task from the view of image restoration. The mathematical model of regularization is analyzed, also the properties of the regularization function with edge preserving. Then, the potential function is chosen and modified according to the tradeoff among the complexity, robustness and the granularity control, and the half-quadratic regularization strategy is applied for the reason of achieving the fast minimization of the energy function. Finally, the specific alternative iteration for the minimization process is described. The proposed method is adaptive to the edge details of images and can reach the minimum values lastly. The experimental results show that the method can improve the subject and visual quality of image. Compared with the MAP based algorithm, it may preserve more image details.

关 键 词：图像压缩 后处理 低码率 正则化 边缘保持 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术]