An Error-Based Statistical Feature Extraction Scheme for Double JPEG Compression Detection

Discovery of double JPEG compression assumes a critical part in digital image forensics. Some effective methodologies have been proposed to identify double JPEG compression when the essential and auxiliary compressions have distinctive quantization matrices. Be that as it may, identifying double JPEG compression with the same quantization matrix is still a challenging issue. In this paper, an effective error based measurable component extraction plan is displayed to tackle this issue. Initial, a given JPEG document is decompressed to form a reconstructed image. An error image is obtained by figuring the difference between the inverse discrete cosine transform coefficients and pixel values in the recreated image. Two classes of blocks in the error image, to be specific, rounding error block and truncation error block, are analyzed. At that point, a set of features is proposed to characterize the measurable difference of the error blocks between single and double JPEG compressions. At last, the support vector machine classifier is utilized to recognize whether a given JPEG picture is doubly compressed or not. Test results on three picture databases with different quality factors have exhibited that the proposed technique can fundamentally beat the best in class system. Moreover as a part of forensic identification, the network is also trained to detect contrast enhancements being made to the image. Mainly three types of contrast enhancement are detected on image i. e. sigmoid contrast enhancement, linear contrast enhancement and gamma contrast enhancement.