The Research On Compressed Noise Modeling And Deblocking Algorithm

A number of image/video compressed standards have been made, most of which involves Block Discrete Cosine Transform (BDCT) and quatization. However, when compressed at low bite rate, BDCT will introduce high quantization noise, which can result in notable discontinuities between image block boundaries. Deblocking algorithms can be used to remove the blocking artifacts and to improve the image quanlity.Deblocking is a reverse problem, and the key point is how to evaluate the intensity and distribution of the blocking artifacts. The quantization step size mainly affects the distribution of blocking artifacts, and also contributions to how to design deblocking artifacts. In this article, deblocking methods are studied with and without knowledge of quantization step size and two different deblocking algorithms are proposed.Without the knowledge of quantization step sizes, a metric of deblocking efficiency is given using image content based on version quality. The metric reacts reasonable to both the discontinuities degree between the image block boundaries and the regressions of the image content. A novel adaptive deblocking algorithm is proposed based on the metric. Experimental results show that the proposed method removes the blocking artifacts while preserving the details of the image content and the deblocking result of proposed method is superior to that of conventional mentods.With the knowledge of quantization step sizes, noisy image is formulated as a combination of original image and compressed noise based on image restoration. Mathematical tools are used to analysis the compressed noise distributions, based on which a noise formulation is given using a noise covariance matrix for image block. And a novel adaptive deblocking algorithm is proposed using non-local means filter based on the noise formulation. Experiments demonstrate that the proposed method outperforms the conventional methods on both subjective quality and objective quality.