Research On Image Block Copy-move Forgery Detection Based On Improved Gray Level Cooccurrence Matrix

Digital images are the most important form of information transfer now.Therefore,many image processing software came into being.The image processing software for the public is very convenient in operation.Only a simple understanding can master how to use this type of software.However,some people use this tool to distort the facts through the second modification of the image,so that most people are deeply affected by it,even involving the safety of personal and property.Therefore,a lot of research directions have been spawned to identify the authenticity of images.The main purpose to find evidence of image modification by looking for modification traces,which can help restore the facts.This has gradually become a hot spot in the field of digital image research.The method of forgery after copying the same image belongs to a method of obtaining authenticity as evidence in passive forensics.Generally,this method will have many areas similar to the original image.For this method of forgery,two methods of detecting feature points and image block similarity are currently the main methods,each of which has its own advantages.In general,the former is better in terms of accuracy and can be used in a wider range.But,there are still many areas for improvement through block matching.This paper chooses the block-based detection method to improve the accuracy of detection and save more time for evidence acquisition.This paper mainly studies the copy-move forgery and detection technology of the same image.The main work includes:(1)When the current block matching detection method is used for image gray feature extraction,which detection effect is reduced and the smooth and natural areas are prone to erroneous detection results.For this problem,an improved feature extraction method based on GLCM and direction measure was developed.First,the direction measure is added to capture the slight changes in the four texture directions of the image;Then introduce weight factors to obtain the micro-directional features of the texture;Finally,the features in the micro-direction and the four Haralick features in the macro are fused to obtain the features with stronger texture description.Experimental results show that the algorithm has strong image recognition ability for the extracted texture features and it also improves the classification accuracy of the texture library.(2)For the problem of poor single feature detection and high time complexity of the current block matching algorithm,the fusion of frequency domain features and space domain features is studied and this analysis method is applied to copy-move forgery detection.First,the frequency domain features and the spatial domain features are fused to obtain stronger features;Then introduced SVD and K-means to speed up block matching;Finally,cardinality sorting is used to locate the forged area.Through a series of test results,it is found that this algorithm can accurately copy and locate a single area of texture.At the same time,it can accurately determine the single forged area of the texture after geometric transformation such as scale transformation and rotation.It is very robust to post-processing such as noise,rotation and scaling.(3)Design and develop an image copy-move forgery detection system.The system can detect whether the image has been forged by copy-move and more accurately locate the copy-move area of the forged image.