Travel grant, Pacific-Rim Symposium on Image and Video Technology (PSIVT), Jan. 2009.
Ph.D. dissertation:
The Valley of Vision: Perspectives and Thoughts of Modeling, Quantizing, and Compensating for Image Processing Improvement Related Studies
Abstract:
The imperfect information received from human eyes reflects the practical situation as known; instead, the perception is modified by personal “tendency.” Digital image processing highly relies on human eyes’ property, if a resultant involves this consideration or thought, even though it is dissimilar to its source from a signal similarity point of view, a better perception quality is still available by the prospective viewers. This dissertation includes three subscopes under this image processing discipline: halftoning, watermarking, and image contrast enhancement, and moreover, totally six individual research works are related. In fact, these works more or less utilize the aforementioned imperfect feature of human eyes for reaching the purposes of giving some further improvements by thoughts of modeling, quantizing, and compensating. Among these, two of the six works, namely co-optimized dot diffusion (DD) and near-aperiodic DD (NADD), attempt improving the printed dot distribution of the DD halftoning technique for yielding a higher visual quality by exploiting the low-pass property of human visual system. Furthermore, three of the six works, namely overall minimal-error searching (OMES), complementary hiding error-diffused block truncation coding (CHEDBTC), and majority-parity-guidance EDBTC (MPG-EDBTC)), consider either halftone patterns or the BTC compressed images as the phenotype of the results after their watermarking processes. These methods lay on the ground of great visual marked image quality to reach either high data capacity or robustness as their purposes. For the last contrast enhancement subscope the parametric-oriented histogram equalization (POHE), it is proposed to greatly reduce the complexity as well as ease the artifact of over enhancing induced by the former local methods. In addition, a modified form of the POHE named correct POHE (CPOHE) is also presented for yielding a higher contrast effect by sacrificing somewhat processing efficiency. Overall, these six efforts are able to be potentially applied on the printing industry, the appeals of secret data transmission and copyright protection, photo editing, or the most attractive subjects of pattern recognition, computer vision in recent years.
Citation:
Yun-Fu Liu, "The valley of vision: perspectives and thoughts of modeling, quantizing, and compensating for image processing improvement related studies," Ph.D. dissertation, National Taiwan University of Science and Technology, 2013.