Improved Tampering Detection for Image Authentication Based on Image Partitioning

Abstract

This paper proposes an improved tampering detection algorithm for image authentication, using distributed source coding (DSC) principle, by partitioning the image into N equal clusters, for Slepian-Wolf coding. The proposed DSC based technique for authentication, models the magnitude variation in pixel domain generated by acceptable distortion and illegal tampering as Normal distributions with small and large variances respectively. DSC coding is done for the feature vectors derived from the partitioned image. The feature extraction algorithm is revamped in the aim to produce feature vectors with statistical features well matched to low density parity check (LDPC) codes, designed for the binary symmetric channel (BSC), as the design of LDPC codes for the BSC is intensely investigated and also a well understood topic. The results reveal that, the proposed scheme's tampering detection is superior than the earlier DSC based techniques.