| Digital image processing, which is connected with computer science, engi-neering, statistics, information science, biology, physics, chemistry and social sci-ence, becomes an indispensable tool for scientific research and social production.Image denoising is the basic tools of image processing, so a good image denoisingalgorithm is so important for the development of this field. The total variationmodel which based on P-M model is one of the classic image denoising algorithm.However, this model makes the denoised image is too vague, affect the image of thevisual effects. For the purpose of protect the edges at the time of denoising, thispaper puts forword two kinds of improved total variation denoising models.Theyare both improved on the TV model.The first is image de-noising algorithms based on spatial correlation filteringand TV model. As we all know, according to decomposit by wavelet, image detailsmainly concentrated in high frequency part, and make the correlation calculationbetween the adjacent wavelet coefficient can increase the accuracy of edge. In thispaper, wavelet coefficient’s correlation calculation is used for control the spread ofTV model, the new model can denoise as well as protect edge details. Three kindsof typically disperse methods are used in the simulation experiments, and the re-suits show that this method can improve the visual effect and enhance the values ofPSNR.The second kind of total variational denoising model is based on morphologi-cal edge detection algorithm and TV model. Mathematical morphology is one of theclassic comprehensive discipline of image processing area. Morphological edgedetection algorithm is a classic edge detection algorithm. This algorithm is get themorphological edge detection algorithm as the weight function of TV model. Thenew model can protect the edgeinformation when denoising. It can improve thevisual effect and enhance the values of PSNR.Those two improved total variational models which suggested in this paper aresimply, small amount of calculation and have a distinct effect for noise image. Those algorithms improve image visual effect, as well as enhance PSNR. Thesimulation results indicate that the two algorithms are all achieve the purpose,which assumed in the time of designed the initial algorithms. |
