Survey On Image Contrast Enhancement Detection

With the development of computer technology and digital imaging equipment,the threshold for operating digital images through image processing software has also become lower and lower,making it difficult to distinguish whether the image has been processed.Moreover,if some processed pictures are applied to improper places,it will cause great harm to society.Image contrast enhancement operation is a common operation to achieve image blurring and retouching.After being used for splicing,copying and pasting operations,image contrast enhancement operation is used to eliminate visual or statistical distortions caused by edges.Image contrast enhancement is a kind of image tampering.Although it cannot significantly change the content of the image,it can blur or highlight some details of the image or change the sensitivity of the image.The detection of image tampering can be achieved by detecting the enhancement of image contrast.Therefore,the detection of image contrast enhancement operation is also a key point of digital image forensics research.In this paper,we firstly analyze the characteristics of the contrast enhancement operation in image tampering,beneficial from the advantages of deep learning in the field of feature extraction,to realize the classification and detection of image contrast enhancement.We first propose a multi-branch neural network to detect the global contrast enhancement operation.By learning the characteristics of the histogram,the network can be well applied to the contrast enhancement detection,especially when the image is JPEG compressed with medium or low quality.After realizing the global contrast enhancement detection of the input image,We explored the local contrast enhancement algorithm based on the classification task and the local contrast enhancement detection algorithm based on the detection task,and realized the detection and localization of the local contrast enhanced image,and both methods achieved good performance.In addition,we also migrated the designed local contrast enhancement detection model based on classification to the image recolor task,and achieved good detection results.