Ranking salient contours for object segmentation
The ultimate goal of Computer Vision is to enable a machine to see and understand an image or scene, at least as well as a human. An important step towards this goal is to partition an image into regions, each corresponding to an object or entity. This is referred to as image segmentation in the computer vision community. Segmentation is an important step towards image understanding and can enhance the performance of many applications such as object detection, object tracking, surveillance, medical imaging, etc. Our framework for Salient Object Segmentation consists of four stages:
1- Finding a set of line segments in the image that represent color or texture discontinuities (edges) 2- Representing the line segments in a sparse graph model 3- Extracting simple closed contours (cycles) in the above graph 4- Ranking the set of closed contours for output
The goal of the proposed project is to research and develop an automated method for ranking object hypotheses (stage 4 in the above framework). Given a set of closed contours found in an image, the problem under consideration is to rank these contours for output, based on how well they bound the salient object in the image. This ranking stage is important since it defines the final output of the object segmentation method. In particular, working with a senior graduate student or postdoctoral fellow, the successful applicant will:
1. Conduct a literature review of the state of the art ranking methods. These methods can include general ranking methods as well as those specifically designed to perform best for top ranked output such as those used for web searches (e.g. Google or Bing).
2. Develop and implement an efficient method for ranking closed contours.
3. Identify the quantity and distribution of data required for training above method, and prepare training data accordingly.
4. Compare ranking results with the currently available (and implemented) ranking method.
Other possible extensions to this project can include:
5. Study of ranking feature, and their inferential power
6. Study of the distribution of contour hypotheses and their power in representing whole objects or parts
1. Good programming skills
2. Good math skills
3. Knowledge of MATLAB programming language
For more information about the lab, visit www.elderlab.yorku.ca.