Research On Filtering Algorithm Of Random Value Impulse Noise

Digital images are often influencd by internal or external factors in process of image generation and transmission.Image noise can cause image quality degradation,and even adversely affect subsequent image analysis and processing.So,image denoising is an important research content of image processing.This paper studies the common impulse noise in images,especially the random value impulse noise.Based on the analysis and summary of the existing main denoising methods,several new method to improve the image denoising performance is further explored.Linear filtering and nonlinear filtering are two typical types of filtering.In this paper,the denoising performances of above two methods are analyzed.The results show that the nonlinear filtering(median filtering)has better effect on removing impulse noise.Based on the median filtering algorithm,several typical improved algorithms are further analyzed,including weighted median filtering,extreme median filtering,and switching median filtering based on sorting threshold(OTSM).The denoising algorithm using geometric structure detection is suitable for removing random value impulse noise.The algorithm uses the image histogram to estimate the image noise rate.The adaptive threshold selection is used to divide all the pixels into three kinds: signal point,noisy point and point to be tested.Then,the geometrical structure is used to perform noise detection on the tested points,and finally the image signal points are reserved,while the median filtering is performed on all pixel values which are determined as noise points.Based on the existing typical image impulse noise filtering algorithm,the following two improved algorithms are proposed to improve the image denoising performance.The algorithm is simulated by using MATLAB,and the performance of the algorithm is analyzed and compared.An improved extreme value median filtering algorithm is proposed.The algorithm detects the noise by the extreme value in the filter window.For the noise pixel,all the pixel values judged to be non-noise points in the neighborhood are used for median operation.According to the detection result,The filter window is adaptively expanded until the window contains non-noise point pixels that are used for the median operation.The simulation results show that the algorithm has obvious advantages in the denoising performance of high-density impulse noise.An improved directional weighted median filtering algorithm is proposed.The algorithm uses image local feature adaptive to set the truncation threshold for detecting noise,and assigns appropriate weights to the noise pixels for median filtering.The simulation results show that the improved DWM algorithm can improve the noise detection rate and strengthen the calculation weight of the non-noise pixels in the neighborhood.For the random value impulse noise with density above 40%,the filtering performance advantage can be obtained.