Date of Graduation
Statler College of Engineering and Mineral Resources
Lane Department of Computer Science and Electrical Engineering
Many lossless image compression methods have been suggested with established results hard to surpass. However there are some aspects that can be considered to improve the performance further. This research focuses on two-phase prediction-encoding method, separately studying each and suggesting new techniques.;In the prediction module, proposed Edge-Based-Predictor (EBP) and Least-Squares-Edge-Based-Predictor (LS-EBP) emphasizes on image edges and make predictions accordingly. EBP is a gradient based nonlinear adaptive predictor. EBP switches between prediction-rules based on few threshold parameters automatically determined by a pre-analysis procedure, which makes a first pass. The LS-EBP also uses these parameters, but optimizes the prediction for each pre-analysis assigned edge location, thus applying least-square approach only at the edge points.;For encoding module: a novel Burrows Wheeler Transform (BWT) inspired method is suggested, which performs better than applying the BWT directly on the images. We also present a context-based adaptive error modeling and encoding scheme. When coupled with the above-mentioned prediction schemes, the result is the best-known compression performance in the genre of compression schemes with same time and space complexity.
Parthe, Rahul G., "Adaptive edge-based prediction for lossless image compression" (2005). Graduate Theses, Dissertations, and Problem Reports. 4183.