Parameter Estimation And Image Recovery For Image Contrast Enhancement In JPEG Domain

With the rapid development of image editing software,the image tampering and forgery has became common.When the image processing techniques are used for the image tampering maliciously,it will generate serious consequences.Thus,image forensics has been drawn more attention by many researchers around the world,many important international conferences and organizations have set up the association or magazine of the researchers for image forensics.Wherein,based on the detecting the content tamper of digital image and the recovery of image blur,the digital content forensics technology is becoming a hot research topic recently in the field of information security.As a branch of forensic technology,image enhancement is divided into pixel domain image enhancement and compressed domain image enhancement,in compression domain image enhancement has low computational complexity,and can effectively balance the compression and enhancement distortion,etc.thus it is a image enhancement tool by tampers.At present,there are many domestic and foreign research results of the image enhancement in pixel domain,few results have been found for the enhanced detection of compressed domain.The main works of this paper:(1)The theoretical analysis a classic image enhancement is carried out in compressed domain algorithm characteristics of the algorithm is proved reversible by mathematical derivation,which supplies a theoretical basis for detecting such enhancements as well as the image recovery.(2)The data characteristics in the frequency domain data is analysied after image compression and enhancement,three methods are designed to detect enhancement and parameter estimation in compressed domain.The first method is based on the periodic distribution of the discrete cosine transform(DCT)coefficients to detect enhancement and parameter estimation;the second is based on the maximum likelihood method to detect enhancement and parameter estimation;the third method combines the previous two approaches.This thesis analysizes the detection results of the enhancement of standard image based on the large number of different compression quantization table,the result is confirmed to be effective by the proposed three methods.The first method enables a fast detection and parameter estimation of the image enhancement,the second method can provide an accurate detection of the image enhancement,the last method combines the advantages of both and achieves a higher efficiency and accuracy.