实时视频压缩中块效应消除方法的研究

REDUCTION OF BLOCKING ARTIFACTS IN REAL TIME VIDEO COMPRESSION

本文较全面的研究了目前存在的各种类型的块效应消除方法 ,提出了一种新型的用于消减视频编码中的块效应的双模式自适应滤波器 (DMA) .同时 ,还提出了基于 MSDS(the mean squared difference of slopes)的客观评估的量化指标 ,它较之 PSNR更逼近块效应问题的本质 ,揭示了主观效果后面的数学实质 .DMA通过分析块边界的像素行为 ,在两种完全独立的滤波模式间做出选择 .它根据 HVS(Human Visual System)特性 ,对平坦区域采用力度较强的平坦区域滤波器实现平滑滤波 .而对细节丰富区域的块间像素 ,采用基于频谱分析的自适应滤波器既能有效消除块效应又能保留图像自身的高频信息 ,从而达到了十分理想的消除噪声的效果 .此外 ,采用了优化策略使 DMA的计算开销降到最低 ,可完全适应实时系统的需求 .

Based on an overall reseach of currently existing various methods for deblocking,this paper proposes a new deblocking strategy named a Double Mode Adaptive Filter(DMA).At the same time,we also propose an objective evaluation exponent based on MSDS,which more adequately exposes the quiddity of blocking artifacts than that of PSNR,and is accordant with subjective evaluation. DMA takes the mode decision between two separate modes by the pixel behavior around the block boundary.According to the human visual systems sensibility to different region, strong filter is applied on the block boundary also inside the block in the flat region mode.At the same time,an adaptive filter based on the analysis of frequency information around block boundaries is applied on the block boundary in other regions without introducing undesired blur.In addtion,a few optimization measures are used to reduce the complexity of computation .So DMA can be used for very low bits video codec in the real time application ,and takes excellent effects.