Historical:SoC2007 project Feature Descriptor

See SoC 2007 overview for usage hints and a page list.

Robust local feature detector and descriptor

Goal: Robust matching of features between multiple images using a Hessian-based detector and a suitable descriptor. A detector and descriptor that takes into account the approximately known distortions will have a much higher matching rate, especially when fisheye or wide angle images are used.

• Implementation of the feature detector and descriptor, and a suitable test suite to verify the correctness of the implementation.

Deliverables

A desired result of the projects would be:

• C or C++ library that implements the detection and description steps.
• An executable for extracting features from image file.
• Test suite to evaluate descriptor on a large amount of images.
• Integration of the library into hugin

Project Schedule

• 0w April 9 - Application selection

• Interim Period: Learn more about the project, the literature that is being used to develop it, and build a testing set with ground truth ~200 images. This can be done asking for pictures to the mailing list.

• 0w May 28 - Begin coding for the project

• 1w June 4 - Create all the evaluation software, (mainly available from the website of Krystian Mikolajczyk)

• 3w June 18 - Implement of the algorithm chosen for the detector of feature points on the image.

• July 9: Students upload code to code.google.com/hosting; mentors begin mid-term evaluations

• 7w July 23 - Implementation of the descriptor to match the feature points detected.

• 10w August 13 - Benchmark and code optimization.

• 11w August 20 - Final report (Google Deadline for all student work),students upload code to code.google.com/hosting; mentors begin final evaluations; students begin final program evaluations.

Required/available ressources

Required knowledge or interest in:

• signal or image processing background
• C or C++ development skills.
• Matlab or octave

Ressources

Literature

Literature about feature detection

• Wikipedia article on SIFT, contains several good references
• Tutorial on local invariant features
• Lecture notes on feature detection/matching

Software

• Framework for testing descriptors

Mentor: Pablo d'Angelo, Herbert Bay, ?

License: LGPL

Students planning to apply

• Pedro Alonso Ferrer