Visual Scene Segmentation

Project: $DO NOT EDIT THIS FIELD$  
Supervisor: Ke CHEN
Difficulty grading: SH/INF=U, BM=N, CM=C
Area:    Artificial Intelligence
Max number of students who can do this project: 1

Visual scene segmentation is a process to find salient and coherent regions from an image, which is the basis for many visual tasks, in particular, for object recognition. Although it is effortless for humans to perform most of such a task, it is challenging for a machine vision system. Many attempts for developing a computational algorithm have been done so far but most of them fail to demonstrate robust segmentation capabilities under general viewing conditions. This project is going to implement a latest image segmentation algorithm, e.g. [1], to evaluate its performance on various types of images against well-defined criteria. The major objectives include

• understanding of a selected algorithm (as well as background knowledge in image processing) and evaluation criteria,
• development of a prototype that implements the selected algorithm and facilitates the access to various types of images for performance evalution,
• design and implementation of a friendly GUI to enable the prototye to be adapted for a real world application with ease.

REFERENCES:

1. E. Sharon, M. Galun, D. Sharon, R. Basri and A. Brandt, "Hierarchy and adaptivity in segmenting visual scenes," Nature, vol. 442, no. 17, 2006, pp. 810-813. (comments: exemplar work for visual scene segmentation)
2. J. Strickrott, "A survey of image segmentation techniques for content-based retrieval of multimedia data," [on-line available]: http://www.cs.fiu.edu/~jstric01/docs/Advanced%20Multimedia%20Databases%20survey_paper.doc (comments: a latest survey for image segementation)

COURSE PREREQUISITES: Image Processing (Computer Vision)