In-camera JPEG compression detection for doubly compressed images

摘要

An illicit photography work can be exposed by its unusual compression history. Our work aims at revealing the primary JPEG compression of a camera image especially when it has undergone an out-camera JPEG compression. The proposed method runs a recompression operator on a given image using a chosen software tool (MATLAB). We measure the JPEG error between the given image and the recompressed version in the Y, Cb and Cr color channels. The in-camera compression can be easily identified by drawing the JPEG error curves. In this paper a simple and high effective method is presented for automatically detecting the compression history of an image. For a doubly compressed image, the proposed method can give the historical compression sequence with the corresponding quality factors and determine whether the first compression is the in-camera compression. Experimental results, carried out on two datasets, show that the proposed method can yield satisfactory detection accuracy, over 96 % accuracy rate for in-camera compression and no false positives with a block size of 512 x 512. The proposed method has universality. It can be applied to multi-compression detection and is robust to different sources of out-camera compression, e.g. Adobe Photoshop. This makes it more practical compared to the previous methods of double compression.