Denoising Algorithm

Digital image is inevitably corrupted by various noises during acquisition and transmission, resulting in degradation of image quality, and bringing serious impacts for follow-up processing of image, such as edge detection, image segmentation, feature extraction, pattern recognition and so on. Therefore, image denoising is a very important link of image preprocessing. In recent years, the image restoration and denoising have attracted widespread attention of people. This paper first introduces the research background and significance of image denoising, and the development overview and evaluation methods of image filtering algorithm; then introduces the classification and mathematical model of image noise, and emphatically introduces the traditional image denoising algorithms: including mean filter , order statistics filter, adaptive filter, morphological filter, the denoising algorithms based on partial differential equations and wavelet transform. On the basis of research for the previous impulse noise filterings, according to the viewpoint of switch filtering, an improved median filtering algorithm based on two-stage noise detection is proposed. This algorithm analyzes and overcomes the shortcoming of the local extremum misjudgement of the max-min noise detector by adding the second noise detection with the local energy of image pixel. The noise pixels are removed using the improved median filtering, while signal pixels hold their gray value and are left unprocessed. Simulation results show that this algorithm has the ability of removing noises and the ability of preserving the partial details of images in comparison with some recent methods especially when the noise density is very high. In order to remove mixed impulse noise and Gaussian noise of the corrupted images, a new filtering algorithm for mixed noise is presented combining median filter and total variational model. The pixels are classified into two sets: impulse noise points and pixels polluted by Gaussian noise. The max-min operator is used first, then impulse noise is detected accurately by using the local energy information of pixels and removed by improved median filter, A detail-preserving adaptive generalized variational functional model is adopted to reduce Gaussian noise contained in other pixels. An edge preserving potential function is selected, which has good robustness to noises and can avoid "staircase" effect of the filtering results, finally a weighted gradient descent flow is developed for image noising with an iterative algorithm based on semi-point scheme. Experimental results show that the algorithm can suppress mixed noise very effectively and preserve image details very well. To improve the performance of traditional wavelet threshold denoising algorithm, A new adaptive threshold denoising algorithm is proposed. First, a new threshold function is produced; Second, a model for wavelet coefficients is presented, in which each coefficient in a sub-band is obeyed to general Gaussian distribution; Third, in order to obtain the adaptive optimal threshold, the variance is estimated from the local neighborhood information of sub-band wavelet coefficients. Experimental results demonstrate that the proposed algorithm has a better peak signal-to-ratio and subjective vision effect, compared to traditional wavelet threshold denoising algorithm.