Block-based adaptive noise filtering for H.264/AVC compression

Noise corruption degrades the R-D performance of H.264/AVC encoders because of a loss of image quality and an increase in bit-rate. Previous noise reduction techniques attempt to avoid the loss of image quality, but they do not take the effect of noise on compression efficiency into serious consideration. In this paper, the noise effect on an H.264/AVC encoder is analyzed and a blockbased adaptive denoising algorithm is proposed. The proposed algorithm applies the denoising operation adaptively to blocks which do not include image details that need to be preserved. Furthermore, the proposed noise filter is not applied when the quantization parameter is larger than a pre-defined threshold because noise is filtered by a quantization operation in the encoder. As a result, both residual and information bits are decreased by the proposed noise filtering operation. Edge blurring caused by the filtering operation is also reduced as a result of the adaptive application of the filtering operation. Compared to cases where filtering is not used, the proposed method shows a maximum PSNR enhancement of 1.98 dB and a maximum bitrate reduction of 41%.