基于纹理方向自适应的H.264去块滤波算法

New Deblocking Filter Algorithm Based on Texture Direction Adaptive Technique for H.264

基于块的混合编码在压缩率较高的情况下会产生明显的块效应,导致编码图像主观质量下降。本文在研究和分析了H.264标准中采用的去块滤波方法后,为了进一步提高去块滤波的性能,提出了一种新的算法。该算法首先利用图像相邻像素的相关性,对帧内预测编码时边界强度的计算进行优化,以降低滤波过程中的计算量。然后根据块边界处图像的纹理方向自适应选择不同强度和模式的滤波算法,从而保证滤波后图像的真实性。实验结果表明:与原H.264的算法相比,该算法能在一定程度上提高重构帧的峰值信噪比(PSNR)值。

Traditional block-based video coders suffer from annoying blocking artifacts when the compression ratio is high, thus the subjective quality of the video images will be decreased. This paper studies the deblocking filter system for H. 264 and proposes a new deblocking filter algorithm to further enhance the performance of the deblocking filter. The algorithm optimizes the computation of the boundary strength for intra prediction exploiting the spatial correlation on an image, then analyzes the texture direction of the block boundaries and choses an appropriate filtering mode to adaptively maintain the authenticity of the reconstructed frame. The experimental result showes that compared with JM8.6 original algorithm, the proposed algorithm can enhance the impersonal quality of the reconstructed image.