|
|
|
|
| 1. Enhanced Reality | |
| Design and implement an "enhanced reality" system, which allows a player to play basketball, golf, or tennis. This system will work by using a projector to display a background scene and related sports objects on a screen, and a video camera to get user pose and actions. You will need to determine the interaction between the user and objects in the scene. | |
| Sample Application: | Video Entertainment, Assisted Learning |
| 2. Signature Verification | |
| Design a system which will verify the authenticity of the user based on verifying the users signature. Use the digital tablet to obtain a signature and verify it with a acquired database of signatures, returning a positive or a negative and a confidence value specifying how well the signature matched to the learned data. This is a continuation of a previous student project. | |
| Sample Application: | Credit cards, Banking security |
| 3. Differentiating apparent contour edges in an edge map | |
| Edges arise from a variety of sources including: occlusion, self occlusion, surface discontinuity or ridges, surface reflective changes, illumination effects such as shadows and highlights, surface texture changes, etc. However, an edge map does not typically differentiate among the sources of an edge. Since apparent contours need to be extracted from among others for successful recognition, it is desirable to identify which edges can potentially belong to apparent contours. There is not much literature on this, so this is a project for the very motivated, highly self driven individual!! | |
| Sample Application: | |
| 4. Determining 3D Shape from 2D Contours | |
| As a 3D scene is projected onto a 2D plane (i.e. an image), 3D regions in the scene give rise to regions in the 2D image, the boundaries of these 2D regions are called apparent contours. These contours contain information about the shape of the original 3D region. This project entails using these contours to reconstruct the 3D shape of the original scene/objects. You will take in a sequence of 2D images, and once the apparent contours are determined, attempt to reconstruct the organization of the 3D scene from which the images were generated from. | |
| Sample Application: | Scene reconstruction |
| 5. Partitioning Visual Form | |
| Attempt to decompose an image into its object components. | |
| Sample Application: | |
| 6. Biometric Recognition: Eyes, lips and face profiles. | |
| From images of the head and face, extract the shape of the lips, eyes, and face profile. Use this shape information to identify the person from a database. | |
| Sample Application: | Security, crime investigation |
| 7. Shape Morphing | |
| Implement a program which given the first and last frame of a video sequence, can interpolate the motion between them. You will use one or more existing shape interpolation algorithms to test on various video sequences, and judge their success by comparison to the actual video frames. | |
| Sample Application: | Movie reconstruction, cartoon animation |
| 8. Image Morphing | |
| The morphing functions in the lab were simple in that they were focused only on pixelwise operations. More intelligent warping methods, including mesh warping, field warping, radial basis functions, etc., prevent errors in intermediate steps by removing double exposure effects, and provide better results due to the inclusion of information from the entire image. You are to implement 3 "intelligent" image warping functions, and test its output of a series of images. | |
| Sample Application: | Movie special effects, image aging |
| 9. Video/Image Tracking/Combining Speech Recognition and Lip Reading | |
| Design an application which tracks the lip outlines of a subject through a series of images, and correlates them to determine if a certain word/phrase was said, and more generally to recognize a new phrase. This entails acquiring training data, and testing the application with test data. The program will need to segment out the region of interest (the lips), fit a curve to the lips, compare the curve progression with the training data and make a decision if the target phrase was spoken. If you achieve a moderate level of success, you can combine the system with a speech recognition system to achieve robust language recognition. | |
| Sample Application: | Spoken Password Verification System, Lip Reading |
| 10. Shape-Based Image Compression | |
| Evaluate 3 separate shape-based image compression algorithms in the framework of teleconferencing. Obtain a set of video sequences, on which you will test the various compression methods. Discern the strengths and weaknesses of each, and obtain the success of each relating to teleconferencing, i.e. speed (frame rate), image quality, compression speed, etc. | |
| Sample Application: | Teleconference |
| 11. Handwritten Character Recognition | |
| Implement a system which takes data from a digital writing tablet and recognizes written words, and the resulting text will be displayed on the screen. | |
| Sample Application: | Digital notebooks, PDAs, |
| 12. ENO Image Interpolation | |
| Implement the ENO interpolation method, and compare its output with other interpolation methods. ENO interpolation is a 1D interpolation method. You will need to make this 2D (can be done in two ways), and test it on a variety of images and resolution changes, and determine the strengths and weaknesses of it versus other interpolation methods. | |
| Sample Application: | Image enlargement |
| 13. Digital Photo Editing | |
| Imagine wanting to remove an old boyfriend/girlfriend from a photgraph. The subject could be removed from the scene, but there would be an empty space left. You will work on a system to restore the background in a realistic manner. | |
| Sample Application: | Entertainment, marketing, blackmail |
| 14. Vision for Robotic Navigation | |
| Design a video tracking system to control robotic servos. Design a robot that can either visually or physically follow a subject around a room. | |
| Sample Application: | Security, service robots |
| 15. Neural Signal Processing | |
| Use shape-based comparison techniques to identify a neuron by the shape of its action-potential waveform. In experiments that record neural signals from a subject, one neural electrode often picks up signals from multiple neurons. Separating out the signals from individual neurons is called 'clustering' and traditional techniques have not used shape information. | |
| Sample Application: | Cognitive neuro-research |
| 16. Structured Light Methods for Shape Reconstruction | |
| Design a 3D scanning system which uses a projector as a structured light source and a video camera to scan an object. The projector emulates a laser light source by displaying a a single line, which you pass over the object and attempt to reconstruct the 3D shape and size of the object. The main hurdle of the project entails setting up and calibrating the camera-projector interaction to be able to determine depth. | |
| Sample Application: | 3D Scanning |