Image And Video Denoising Based On Noise Level Estimation

With the rapid development of computer technology and network rapid popularization, images and videos have become an important part of our life. At the same time, we look forward to higher quality of images and videos. However, Image and video produce all kinds of noise with various reasons, All of these provide great impulsion for digital image processing technology.The paper mainly studies the noise level estimation and block-match images and videos denoising based on noise estimation, describe Block Matching and3-D Filtering in detail. Finally, we point out the defect and improve it. The paper mainly includes the following contributions:(1) According to the CCD digital camera imaging pipeline and image noise model, we propose a new noise simulation method which make the synthetic noise more close to natural noise. At present most of the denoising algorithms often assume the noise is additional white Gaussian noise (AWGN). However, the real noise produced by CCD digital camera is not simply additional which is closely related with image RGB values. Because there are many nonlinear processes, such as white balance and gamma correction are embedded in CCD camera.(2) We present a new filter-bimedian filtering which is used to estimate the noise variance and noise level function. At present most of the denoising algorithms often assume the noise variance is known. However, the noise variance is key parameters of image denoising algorithm.(3) For inhomogeneous noise, we present an adaptive local bilateral fiter to remove it. At present, the noise is assumed to be homogeneous in denoising algorithm. In fact, the inhomogeneous noise is objective existence. Based on the assumptions of local noise consistency, we segment the noisy image by the improved Graph-Based image segmentation. To estimate the noise of each region, we remove the noise effectively by the different parameter of noise.f(4) For the state-of-the-art Nonlocal Mean and Block Matching and3-D Filtering, we estimate the denoising parameter by noise level estimation. Meanwhile, we design a strategy of fast block matching, make the NLM and BM3D algorithm be applied to video denoising according to reduce the time complexity from o(n2) to linear o(n).