Loading...
Points, patches, edges, contours, lines, vanishing points, and segmentation techniques.
How do we find corresponding points between two images? This classic problem underlies image stitching, 3D reconstruction, object tracking, and more. The solution: find distinctive features and match them.
What is this chapter about? We study how to detect interesting points (corners, blobs), describe them in a way that's robust to changes in viewpoint and lighting, and match them between images.
Why does this matter? Feature matching is the foundation for:
How the topics connect: We start with feature detectors that find interesting points—corners, blobs, edges. Then feature descriptors encode the local appearance around each point. Finally, we learn matching strategies that robustly pair up corresponding features despite noise and outliers.
Click any topic to jump in
Harris, SIFT, FAST, and ORB — finding distinctive points like corners and blobs that are repeatable across views.
Gradients, Sobel, Canny, and LoG — finding boundaries and contours through intensity change analysis.
SIFT and BRIEF descriptors, ratio test, and cross-check — encoding local appearance for robust matching.
Accumulator-based voting for lines and circles — detecting structured geometric shapes from edge pixels.
This chapter is part of PixelBank Premium. Create a free account, then upgrade to read the full lesson — concepts, walkthroughs, and exercises.