CS 143 / Project 2 / Local Feature Matching

Edges

In the process of implementing the Harris detector, I found the edges of the image with gradients in the x and y direction, then thresholded and maximized the highest points to find interest points.

Next, I implemented a pseudo-SIFT function, which found the orientation gradients around given interest points, in order to create orientation histograms and feature descriptors.

Points

When matching features across images, I implemented a Euclidean distance bound, such that matching points had to be within a variable number of pixels of each other, in addition to doing the nearest neighbors ration test. However, somewhere between the feature detector and the feature matcher, something went wrong, and the program is not matching the majority of the points successfully.

Example of code with highlighting

The javascript in the highlighting folder is configured to do syntax highlighting in code blocks such as the one below.

Results on Given Image

Clearly, with four matches out of 60, this is not nearly as good as it should be; I have an error in my code somewhere.