Bioinformatics-inspired analysis for watermarked images with multiple print and scan

• Main Author: Garhwal, Abhimanyu Singh (Author)
• Other Authors: Yan, Wei Qi (Contributor), Narayanan, Ajit (Contributor)
• Format: Others
• Published: Auckland University of Technology, 2018-03-28T02:23:24Z.
• Subjects: Multiple print and scan; Multiple sequence alignment; Local pairwise and global alignment; Image quality metrics; Image analysis; Pattern matching; Phylogenetic tree; Bioinformatics tool; Thesis
• Online Access: Get fulltext

 Image identification and grouping through pattern analysis are the core problems in image analysis. In this thesis, the gap between bioinformatics and image analysis is bridged by using biologically-encoding and sequence-alignment algorithms in bioinformatics. In this thesis, the novel idea is to exploit the whole image which is encoded biologically in DNA without extracting its features. This thesis proposed novel methods for identifying and grouping images no matter whether having or not having watermarks. Three novel methods are proposed. The first is to evaluate degraded/non-degraded and watermarked/non-watermarked images by using image metrics. The bioinformatics-inspired image identification approach (BIIIA) is the second contribution, where two DNA-encoded images are aligned by using SWA algorithm or NWA algorithm to derive substrings, which are exploited for pattern matching so as to identify the images having a watermark or degradation generated from MPS. The outcomes of identification affirm the capability of BIIIA algorithm. Furthermore, it asserts that DNA-based encoding is the best way for digital images as well as SWA algorithm is the best one for the sequence alignment. The last one is the bioinformatics-inspired image grouping approach (BIIGA), where the DNA-encoded images are aligned by using multiple sequence alignment (MSA), which is exploited by using the phylogenetic tree to group the watermarked / non-watermarked and degraded / non-degraded images; the resultant analysis confirms the potential of BIIGA algorithm. All three methods are empirically verified and validated by using real datasets.