基于自适应邻域的空域错误掩盖算法

Algorithm of Adaptive-Region Based Interpolation for Spatial Error Concealment

针对以固定邻域内的像素为参考进行图像错误掩盖时,往往容易导致边界信息的丢失或产生阴影和虚假条纹的问题,提出了根据错误块邻域图像特征选择自适应参考像素集的空域错误掩盖算法.介绍了一种有效的分割方法,将邻域分割为与错误块相关和无关的两个区域,并以相关区域内的像素为参考进行空域内插.实验结果表明,该算法掩盖的视频图像既能有效地重建多个方向的边界信息,又能抑制或消除无关区域的影响,可广泛应用于掩盖MPEG、JPEG和H.26x等基于块编码码流中的错误.

Spatial error concealment using pixels in fixed neighborhood of error block may result in edge blurring, shadow and false textures. This paper proposed a novel algorithm, which can conceal the error block by spatial interpolation using pixels in adaptive-region. Firstly, an effective region partition algorithm is introduced and all the neighborhood blocks are partitioned into two parts: independent region and correlative region. Then the error block is concealed by pixels in the correlative region. The experimental results show that the image concealed by the proposed method can restore major edge of several directions and restrain the diffusion of independent region. The proposed algorithm is applicable to errors in MPEG, H.26x and JPEG coded images or videos transmitted over error-prone networks.