Thepade’s Sorted Block Truncation Coding Applied on Local Binary Patterns of Images for Splicing Identification Using Machine Learning Classifiers

The era of digitization has accelerated communication and information sharing immensely. With ever-growing digital advancements in technology and applications cybersecurity poses to be a pressing issue. The amount of growth in data exchange is exponential thus making automated processes a vital tool to deliver security. Image editing technologies manipulate image data and have enabled all types of users to tamper images resulting in widespread fake images. Distorted information carries heavy consequences and thus a reliable image forgery detection system is essential. This paper proposes a machine learning-based approach for image splicing detection using the global and local characteristics of the image. TSBTC N-ary, with the value of N = 12,14 and 16, is applied along with LBP for feature extraction and various Machine learning classifiers are implemented and compared for image splicing detection. The performance of the proposed method is tested and validated on 3 benchmark datasets: CASIA V1 Dataset, Columbia Dataset, and Columbia Uncompressed Dataset. Results are evaluated based on various performance metrics.