MIT & MITM 12/18cp Projects
Is the fractal dimension competent for description of textured images?
This project will provide a holistic investigation of the abilities of various fractal dimensions and multifractal dimensions to segmentation and classification of textured images.
The fractal theory is a powerful mathematical tool for modelling many natural phenomena, which have high complexity and irregularity and can not be well treated in general by Euclidean geometry. Since Pentland demonstrated that most natural surfaces are spatially isotropic fractals and their intensity images also present some fractal properties, fractal estimation has been extensively applied to many fields of digital image processing. However, due to the limitation of quantification and spatial resolution, most digital images are not idea fractals, but merely semi-fractals with anisotropic and inhomogeneous scaling properties. As a result, a great variety of approaches have been developed to approximate the fractal and multifractal dimensions of digital images.
This project aims to make a holistic and comprehensive investigation of texture features derived from various fractal models and multifractal models, evaluating their abilities to differentiate natural textures through comparing their performance in segmentation and classification of textured images. The output of this project will facilitate natural image analysis, retrieval, and understanding.
By finishing this project, students will learn some technologies in digital image processing and multimedia and be more experienced in programming and problem solving.
Keywords: Texture image analyse, Fractal dimension, Feature Extraction
Supervisors: Prof. David Dagan Feng, Dr. Yong Xia
Research Locations: Level 3, School of IT building J12