Research On Color Image Splicing Localization Based On QPCA Noise Inconsistency

In recent years,the study of splicing image localization has been a hot topic.Domestic and foreign research teams have proposed many high accuracy localization schemes by studying the inconsistencies between the original region and the splicing region.The existing localization approaches mostly ignore the color information and relative information of image channels,resulting in the feature describing the image information are insufficient.Hence,the existing localization schemes achieve low accuracy.To address the above issues,this paper proposes two effective splicing localization algorithms of color image.The main work is as follows:1)To make full use of the color information and the inherent relationship among the three channels of color image,this paper proposes color image-spliced localization based on quaternion principal component analysis and quaternion skewness.In order to avoid the problem of information loss,the paper designs a feature extraction algorithm based on QPCA,which take the color information of color image and the correlation among RGB three channels into consideration,and enhances the difference of classification features.In order to compromise the accuracy and time complexity of the localization scheme,this paper adopts a coarse-to-fine segment strategy.The specific steps of the proposed scheme are as follows: Firstly,the image is divided into 64 × 64 size image blocks;Secondly,all image blocks are divided into two clusters to obtain the initial classification results according to the noise features extracted by QPCA,and the edge blocks in the initial classification are searched.Thirdly,the image is further divided into smaller image blocks with 32 × 32 size to obtain the second classification.Finally,the second classification is performed as a final classification of the boundary blocks in the initial classification.Experimental results show that the proposed scheme is superior to the existing positioning scheme,and achieve high localization performance.2)To address the problems of many pixels incorrectly detected and splicing edge is discontinuous in the existing localization approaches,this paper proposes a splicing localization optimization scheme based on morphology and quaternion hue.Under the premise of noise inconsistency,this paper proposes a splicing localization optimization algorithm based on morphology and feature similarity,which can effectively reduce the phenomenon of pixels incorrectly detected.The paper proposes splicing boundary optimization algorithm based on quaternion hue to improve the performance of the splicing edges localization.The quaternion rotation theory and color center of gravity model can effectively locate the pixels of splicing boundary,which further improves the localization accuracy of splicing region of color images.