Prof. Roland Wilson,  Dr Nasir Rajpoot,  Yu Song

Stereo vision is an effective way to perceive depth and reconstruct object surfaces from perspective images. Some stereo applications include terrain mapping, depth measurements and robot navigation. Although depth perception is not difficult for humans, there are still many issues with inferring depth information from stereo images despite of two decades of research. The proposed algorithm employs a multiresolution coarse-fine strategy for disparity estimation. From the preliminary results of synthetic data and real data, the proposed algorithm has demonstrated the potential to tackle issues like small objects and occluded pixels.

• Computer Vision, Stereo Vision,
• Multiresolution Coarse-Fine Strategy,
• Disparity Estimation ...

Dr Nasir Rajpoot,  Yu Song

In this project, we examined two recent approaches, Independent Component Analysis and Refiltering, to achieve audio source separation. This project included studying theory and algorithms behind these two approaches, conducting various experiments to demonstrate separating audio sources, discussing evaluation metrics and evaluating these two approaches. Some novel applications were also introduced.

This is part of Yu Song's third year (final year) undergraduate research project.

• Audio Signal Analysis, Source Separation,
• Blind Source Separation (BSS), Independent Component Analysis (ICA),
• One Channel ( One Microphone ) Source Separation, Refiltering,
• Novel Applications,
• Java, C, Java Native Interface, Matlab ...

More information: http://www.dcs.warwick.ac.uk/~yu/ug/projects.html