Authentication Method for Splicing Manipulation with Inconsistencies in Color Shift

Splicing manipulation is a common way for image forgeries,according to the difference of color shift between splicing areas and original part,we put forward a kind of auto-detection method for splicing images based on estimation of color shift.First of all,the suspicious image is divided into small sub-blocks,size of which are n×n.After that an improved average color deviation calculation method is applied to estimate color shift of each sub-blocks.And then the collection of sub-blocks is set in radial directions of top,left,right on the suspicious image as the reference region,with scale selection of edge sub-blocks,color shift between each sub-block and reference is calculated.At last,the splicing area can be localized automatically by the comparison of color shift distance with empirical threshold.As a result,the inconsistency of color shift in a splicing image can be exposed after the localization of the splicing area.Experiments show that the proposed method can detect splicing images and localize splicing area automatically,and provide a scientific and quantitative evidence for the detection of splicing manipulation.