A number of simple computer vision projects
Two different methods for aligning three channels of an image inorder to make an RGB image. This project takes digitized Prokudin-Gorskii glass plate images and automatically produce a color image with as few visual artifacts as possible. The alignment methods are either through gradient descent or through exhaustive search for finding the alignment with the smallest error.
This project is designed to detect and visualize blobs in grayscale images using a multi-scale Laplacian of Gaussian (LoG) approach. The program reads an input image, builds a scale-space using Laplacian filters, detects blob-like structures, and finally visualizes them by drawing circles around detected blobs on the image.
This project stitches two images together based on keypoint matching between two images. This can be used to create panoramic images from multiple images.
A simple implementation of reconstructing shape from different shading. This implementation uses different integration methods that can be compared
Implementation of the Tomasi and Kanade affine structure from the motion method with Carlo Tomasi’s 101-frame hotel sequence.