低码率视频的小波域软阈值去方块滤波算法的研究

本文针对低码率条件下基于子块DCT的视频压缩产生的方块效应,提出了一种小波域上进行软阈值滤波的改进算法。新算法根据图像的平坦区和单调变化区等非纹理区易于通过平滑滤波去噪的性质,以非纹理区作为参考对象计算阈值,并且采用虚拟滤波的思想简化了阈值的计算,同时根据低码率压缩视频的特点修改了在低层子带的软阈值处理过程。对测试序列Susan的实验结果表明,新的算法不仅获得很好的主观去方块效果,而且可以对帧内和帧间编码模式平均分别获得约1.0dB和0.28dB的客观质量(PSNR)的改善,因此新的算法适用于可视电话等低码率视频的应用。

Study on wavelet-based soft-thresholding deblocking algorithm for low bit rate video

This paper presents a new wavelet-based soft-thresholding deblocking scheme for low bit rate video, which is compressed by block based DCT. According to the fact that the non-texture areas can be denoised easily with smoothing filtering, the new algorithm uses these areas as reference in estimating thresholds for each of the high subbands. It adopts the virtual filtering to simplify threshold's computation, and modifies the thresholding at the lowest level according to the characteristics of low bit rate video. Experiments to the test sequence Susan show that the new algorithm can not only obtain satisfactory subjective quality, but also get PSNR gains, which are, on average, about 1.0 dB for intra coding and 0.28 dB for inter coding. Therefore, the new algorithm is suitable for low bit rate video applications, such as videophone etc.