Term Project

Term project topics for Fall 2016 have been posted here. You are supposed to submit a report on your selected topic and the implemeted code. Please contact me before submitting your report.

The deadline to choose a topic is December 5.

The deadline to submit your report is December 20

The project teams can include two students at most.

=========================================

Term projects are either research projects about one of the subjects from the first list below or an implementation of one of the algorithms from the second list. If you choose a research project, you will need to submit a report and present it at the end of the semester (the dates will be announced later). You may also work in groups of two students.

In case you have any other suggestion for project topic you may contact me.

List 1: Research Project Topics:

•  Kalman filter is used for tracking moving objects. Prepare a report about Kalman filter and how it wroks.
• Structure from motion and 3D reconstructing a scene from the video.
• Object matching and recognition
• Content based image retrieval
• Wavelet Transform based segmentation

List 2: Implementation Project Topics

• A program to add random noise to an image and remove the noise afterward. Try different types of added noises and compare the image after noise removal with the original image. Discuss your results. Your project should include at least one method for evaluating the performance of each noise removal algorithm.

• Develop a program to segment an image based on the pixel colors (use RGB images). Define your metric to compare the similarity of color values. Make your algorithm independent from lighting.

• Apply Fourier transform to different gray level images. Compute the power spectrum and discuss the peak points of the spectrum. Change the value the peak points in power spectrum (to zero for instance), apply inverse Fourier transform and discuss the result.

• Use generalized Hough transform to detect an arbitrary shape in an image. Use edge image to extract the probable locations of the object in the image. Extend the algorithm by making it less sensitive to the minor changes in the location of the boundary pixels.