Image splicing detection system using intensity-level multi-fractal dimension feature engineering and twin support vector machine based classifier

Electronic images have become an essential origin of information nowadays, the authenticity of images has become important. Several techniques used for forgery have come into existence like an intrusive method and non-intrusive method. Identification of image forgery is becoming more challenging day by day because of advancements in the processing of electronic images. Consequently, Image forensics is the core part of security applications designed to restore digital media loyalty and acceptance by revealing various methods of counterfeiting. The suggested work compares different feature extraction methods for forged images for the identification of spliced images. In classification techniques, a very difficult issue is to choose features to differentiate between classes. Various features are extracted from real and spliced images with the help of a method dependent on a spatial Gray level like Gray-Level Run Length Matrix etc. This paper describes the methodological considerations involved with the concept of multifractal analysis and with this its emphasis on the Differential Box-Counting method for fetching the Intensity-Level Multi-Fractal Dimension. Further, the research paper evaluates different state of the art in image splicing techniques, and it has been observed that Twin Support Vector Machine as classifier achieves significant efficiency with Intensity-Level Multi-Fractal Dimension as Feature extraction method as compared to other methods.

Automated summary

This paper investigates image forgery identification methods, compares different feature extraction methods and classification techniques, and finds that Support Vector Machine with Intensity-Level Multi-Fractal Dimension as the feature extraction method achieves significant efficiency in image forgery identification.