A Study On The Filtering Algorithm For Adaptive Extended Window Based On City-block Distance

Today, with the development of the network and communication technology, imagesignals carrying information have been widely used in many fields of scientific research,industrial and agricultural production, military technology, health care and education.However, in the process of image signals’ forming、transmission and receiving, image signalsare disturbed by acquisition system, transmission media and imaging system, etc. So theimage can introduce different levels of noise, making the image’s quality decline and affectingits visual effects. Since many of the subsequent image processing work(such as edge detection,pattern recognition, image segmentation, etc) are largely dependent on the effect and qualityof noise removal, effective denoising work is a very critical aspect in the image processing.With the advancement of image processing technology, in order to achieve the purposethat both preserving image details and effectively removing the noise, people proposed anumber of filtering algorithms on the basis of the common filtering algorithms―the meanand median filtering algorithm. In this paper, a lot of denoising methods in the previous imageprocessing were studied. Aimed at the characteristics of the gray image, combining thecity-block distance and the principle of correlation between adjacent pixels is applied to theimage denoising work, and achieves good results.This paper first describes the research background、the meaning and research status athome and abroad. The research contents through understanding image denoising technologywas expanded correspondingly; Then, the types of noise and evaluation criteria of imagequality were described, the traditional digital image processing methods and their advantagesand disadvantages were discussed; Last, on the basis of the introduction of digital imagecommon distance measure function, combines the principle of correlation between adjacentpixels, and builds the window extended adaptively according to the city-block distance, thusput forward adaptive extended filtering algorithms based on city-block distance. Thealgorithms can improve the peak signal to noise ratio, remove the noise effectively andpreserve image details.This research work is mainly concentrated in the following respects:(1) The paper first analyzes the characteristics of common distance measure function as the window extension mode; Then, choose the city-block distance as window extension modeto make it extend faster and the program implement easier; Last, put forward adaptiveextended filtering window based on the city-block distance. The window can adjustadaptively according to the number of signal points within it, and the size of the window is nolimitation. The paper proposes adaptive weighted mean filtering algorithm based on city blockdistance, especially for the larger-noise-density image, denoising effect is more significant.(2) Given the correlation between the adjacent pixels of the image, we hope to use thefull range of information surrounding pixels in the process of filtering. Based on thisconsideration, the paper puts forward adaptive overall situation filtering algorithm based oncity-block distance. The algorithm can guarantee the current point as coordinate origin, andthe signal points in the four quadrants around the point are all available.(3) In the process of filtering, the distance between the current point and signal points isinversely proportional to the contribution of the signal points to the current point. That is tosay, when the larger distance between the current point and signal points is, the smallercontribution of signal points to the current point is. So the paper proposes improved adaptiveoverall situation filtering algorithm based on city-block distance. Regarding the current pointas the coordinate origin, and the Euclidean distance of signal points to the current point as theweights in the four quadrants, the pixel values of gravity position can be got by weightingaverage the signal points. And the adjustment of the contribution of the current point to thesignal point can be achieved through weighting average the pixel values of gravity position.In order to verify the validity of the algorithms, the images with different density noisewere tested under the programming environment of VC++6.0. Filtering results werecompared with median filter、median filter and IAMF algorithm from two aspects of filteredvisual effects and objective performance evaluation criteria. Simulation experiment resultsshow the algorithms can effectively remove the noise and keep image detail. One ofthem―adaptive weighted mean filtering algorithm based on city-block distance, especiallyfor the larger-noise-density image, denoising effect is more significant.