Shape-simplifying image abstraction

• Main Author: Padmavathy
• Other Authors: Deepu Rajan
• Format: Final Year Project
• Language: English
• Published: 2011
• Subjects: DRNTU::Engineering::Computer science and engineering::Computing methodologies::Image processing and computer vision
• Online Access: http://hdl.handle.net/10356/46380
• Institution: Nanyang Technological University

The report documents a final year project at the School of Computer Engineering. In this project, a thorough study of algorithms for image abstraction is conducted to lay a foundation for the design and development of the Mean Curvature Flow (MCF) based framework. The framework demonstrates the constrained MCF technique, in an iterative and incremental manner, to simplify the image content. This image simplification technique – which combines the constrained MCF and Shock filtering processing (to effectively produce a stylistic abstraction of an image) – is presented in the research paper ‘Shape- Simplifying Image Abstraction’ [1] by Henry Kang and Seungyong Lee. Image abstraction refers to the task of simplifying scene information in the image while retaining or emphasizing meaningful features to convey. A lot of research work have been conducted to study and invent image abstraction techniques such as image segmentation, curve fitting, mean curvature flow etc. The technique chosen to study in this first part of the project is an integral method which uses constrained MCF for simplifying image content (shape and color) and shock filtering to protect important structures in the image (shape boundary, edge). The study is conducted in two steps. Firstly, the combined MCF and shock filtering algorithm is investigated thoroughly to compare with other image abstraction methods, and identify the limitations. Secondly, the proposed framework for the improved MCF based method in the paper [1] is studied and analyzed carefully for design and development in the second part of the project. At the end of this part, the MATLAB program is implemented for the purpose of experimenting and comparing MCF method with existing methods for image abstraction. In the next part of the project, the constrained MCF based framework is implemented in MATLAB to demonstrate the iterative image evolution process. In this framework, the constrained MCF process is running in iterative manner and uses the smooth vector as a constraint to protect local image feature. The framework allows user to control the number of iterations – the level of abstraction – as well as the image area to be protected – not adversely affected by the simplification process. The motivation for the development of this framework is to fully understand the novel techniques for image simplification based on MCF and to provide a necessary tool for any further research of related MCF based technique.