A Study On The Application Of Interpolation ELA Algorithm In Image Denoising

With the rapid development and broaden applications of image technology.image processing has played a crucial role in the field of computer vision. However,due to various factors (including defects of the external environment and hardwareequipment etc), the image will suffer noise pollution in the process of acquisition,transmission and processing. Polluted images not only affect people′s visual effectbut also reduce the amount of information they convey. It will bring difficult toimage post-processing work. Image denoising is a prerequisite for image processingwork.This thesis studies for salt and pepper noise. We apply the classic videointerpolation ELA algorithm to remove salt and pepper noise and propose IELADalgorithm and MELAD algorithm.The main results of this paper are as follows:(1) According to the edge-preserving demand of image denoising and based onthe similarity between image information and video information, this thesisintroduces video interpolation ELA algorithm to image denoising. The noise pointsin the image are considered as unknown interpolating pixels in the video. We use thepixels which are in the ELA window to process noise points by ELA algorithm.Under VC++6.0programming environment, we verify the effectiveness of the ELAalgorithm for image denoising;(2)According to the deficiency of ELA algorithm in the processing ofhorizontal direction of the video image and combined with the characteristics ofnoisey image, we improve the ELA algorithm and propose Improved Edge-basedLine Average Denoising algorithm. In this algorithm, an extended5+5ordervertical-window is added based on the5+5order horizontal-window. First, weshould give a preliminary judgment for the direction of the noise point. Then,according to the direction, using the both windows. we calculate the directiondifferences and select the direction whose direction difference is minimum as the best filtering direction. Finally, we use signal points in the best filtering direction torecover the noise point;(3) Traditional ELA algorithm has a limited restoration of directions. For theproblem, Multi-directional Edge-based Line Average Denoising Algorithm isproposed. First, to avoid the influence of other noise points in the image on filteringresults, we process the noisey image by switching median filter. Then, we calculatethe differences of the five directions in5+5ELA window and then find the directionwho has the minimum difference; Secondly, We should find a smaller direction whois adjacent to the smallest direction and has the smaller difference; Finally, We usethe weighted average of signal points’ gray values in the both directions to recoverthe noise. This algorithm allows the recovery directions expand from five directionsto any directions from22.5degrees to157.5degrees.In order to verify the validity of the algorithm, the images with low density andthe images with high density are filtered under the programming environment ofVC++6.0. Filtering results are compared with the Standard Median FilteringAlgorithm and the Improved Adaptive Median Filtering Algorithm from two aspectsof filtered visual effect and objective performance evaluation criteria. Simulationexperiment results show that IELAD algorithm can effectively remove the noise andkeep image details to some extent and MELAD algorithm can achieve better filteringeffect for low-density noisy image.