Hybrid Denoising Algorithm Based On Image Filtering

With the development of computer technology, digital image processing technology has been widely concerned. People also put forward a higher requirement for quality of the image. With the process of generation and storage, it is hard to avoid some amount of random noise which makes the quality of the picture is poor. How to eliminate the noise and improve the quality of the image becames a hot issue of research. This paper studies the image denoising algorithm based on filtering. It mainly includes the following three aspects:First, it summarizes common types and characteristics of the noise, describes the spatial filtering noise and the relevant basic knowledge. Based on several classic image denoising filtering algorithms such as median filtering, wiener filtering, average filtering and the non local means filtering, it uses the mean square error, peak signal to noise ratio, the denoising time and details remain for analysis, which we can find that the image of median filtering, wiener filtering, average filtering method is blurry, and it can't keep the details well,so the quality is poorer than the non local means filtering whose image quality and details characteristics are clear. But the algorithm is inefficient.Second, based on the non local means filtering algorithm it proposes a hybrid image denoising algorithm on filtering. The algorithm is based on markov chain- monte carlo random sampling according to the self-similarity of natural images for sampling. It makes a limited number of iterations after get a stationary markov chain which can get the approximate samples as required to replace the traditional non local average filtering algorithm through each point in the convolution so as to raise the operation efficiency.Third, in the process of image denoising, improving the efficiency of the operation is bound to lose certain signal-to-noise ratio.To improve the SNR of images, it needs to carry on the improvement of weight.There are several kinds of weights applied in image such as denoising function, exponential function, cosine function, based on the robust estimation of the weight function. Experiments for the comparison and analysis of their advantages and disadvantages are done, which in turn, this paper proposes a new weighting function. Making the improved algorithm more efficient in operation with keeping high level of signal-to-noise ratio.