Effects of Image Compression on Face Image Manipulation Detection: A Case Study on Facial Retouching

In the past years, numerous methods have been introduced to reliably detect\ndigital face image manipulations. Lately, the generalizability of these schemes\nhas been questioned in particular with respect to image post-processing. Image\ncompression represents a post-processing which is frequently applied in diverse\nbiometric application scenarios. Severe compression might erase digital traces\nof face image manipulation and hence hamper a reliable detection thereof. In\nthis work, the effects of image compression on face image manipulation\ndetection are analyzed. In particular, a case study on facial retouching\ndetection under the influence of image compression is presented. To this end,\nICAO-compliant subsets of two public face databases are used to automatically\ncreate a database containing more than 9,000 retouched reference images\ntogether with unconstrained probe images. Subsequently, reference images are\ncompressed applying JPEG and JPEG 2000 at compression levels recommended for\nface image storage in electronic travel documents. Novel detection algorithms\nutilizing texture descriptors and deep face representations are proposed and\nevaluated in a single image and differential scenario. Results obtained from\nchallenging cross-database experiments in which the analyzed retouching\ntechnique is unknown during training yield interesting findings: (1) most\ncompetitive detection performance is achieved for differential scenarios\nemploying deep face representations; (2) image compression severely impacts the\nperformance of face image manipulation detection schemes based on texture\ndescriptors while methods utilizing deep face representations are found to be\nhighly robust; (3) in some cases, the application of image compression might as\nwell improve detection performance.\n