Modeling of visual patterns

• Main Author: Chu, Xinqi.
• Other Authors: Chan Kap Luk
• Format: Final Year Project
• Language: English
• Published: 2009
• Subjects: DRNTU::Engineering::Electrical and electronic engineering::Electronic systems::Signal processing
• Online Access: http://hdl.handle.net/10356/15816
• Institution: Nanyang Technological University

Natural images reveal an overwhelming number of visual patterns from objects and scenes in nature. Human vision system identifies and recognizes scene images or ob- jects in the images based on their visual patterns. Modeling these visual patterns is of fundamental importance for generic vision tasks, such as perceptual organization, segmentation, and recognition. Successful implementations of these tasks can enable applications in medical imaging, video surveillance, media analysis, human computer interaction and many other interesting fields. In literature, visual patterns have many different sets of modeling methods. This report delves into both descriptive and gener- ative methods where each method has different emphasis on visual pattern modeling. One of the descriptive model, Gabor filter banks, is investigated for the modeling of texture patterns in chapter 2 because Gabor filters responses are believed to be able to obtain sufficient statistics which best describe a texture pattern. Furthermore, a novel strategy to make this model rotation and scale invariant is proposed in this chapter. The generative model for the modeling of local visual patterns is investigated in chapter 3, where the emphasis is on the graphical models which describe the topology of the components of a complex probability model, clarify assumptions about the rep- resentation, and lead to algorithms that make use of the topology to increase speed and accuracy. Therefore in chapter 3 the goal is to find a generative model that is the best fit to the data rather than extracting the most representative statistics as that in chapter 2. A new way of modeling a class of local patterns by summarizing a collection of images of the visual pattern is proposed in this chapter. By exploring global texture patterns and local visual patterns with their respective modeling methods, this report gives an introductive overview about what is a visual pattern, what is the difference between global texture pattern and local visual pattern and how can they be modeled in different ways. On top of that, two new modeling methods for different types of patterns are presented in this report.