Sharpening Detection Algorithm Based On Sparse Coding

With the development of computer technique and the expansion of network bandwidth,it's getting easier to spread media stream like images and videos.Those digital images have been widely used in various industries,e.g.medical,education and news report.Anyone can easily edit or tamper images with image editing software.If some people spread tampered images for their own interest,it may cause rumors,ruin the reputation of others and have a bad influence on the stability of society.This paper focus on unsharp mask detection.Image sharpening is an operation to filter the edge of an image and add the strengthened edge to the original image.From previous research,sharpening will cause overshoot artifacts and bring rise and falls around edge area.There were two main sharpening detection categories.One is to model the edge area and the other utilizes local textures to achieve statistic features.Those two categories above both apply manmade features.In this paper,two sharpening detection methods that extract features automatically are proposed.The main process starts with the extraction of edge points.Then a block centered with edge point is chosen to do normalization and differencing.Each block is turned into a one dimensional vector.The first algorithm applies K-SVD to extract dictionary and the second algorithm utilizes LC-KSVD to do the dictionary calculation.Sparse coefficient is calculated based on the dictionary above.Also,in order to reduce dimension of the extracted feature,max pooling is applied.Finally,SVM classifier is utilized to distinguish sharpened images from unsharpened ones.After experiments,algorithms in this paper increase the accuracy of detection by 6% when the strength of USM sharpening is weak.