Computer vision projects. Most of the algorithms are programmed from scratch.

• (Final project) Stereo visual odometry

  • Feature detection and matching
  • Stereo to depth map (Disparity map)
  • Lowe's ratio test
  • PnP RANSAC for motion estimation (Transformation matrix)

• (mp1) Image Segmentation

  • Sequential Connected Component Labeling
  • Size/Noise Filter

• (mp2) Morphological Operators

  • Dilation
  • Erosion
  • Opening
  • Closing
  • Boundary

• (mp3) Histogram Equalization

  • Improve contrast and brightness

• (mp4) Histogram-based Skin Color Detection

  • Training a 2D histogram in HSV color space for Skin Color Detection

• (mp5) Canny Edge Detector From Scratch

• (mp6) Hough Transform

  • Hough Transform algorithm for straight line detection

• (mp7) Template-matching based Target/Motion tracking

  • SSD (Sum of Squared Difference)
  • CC (Cross-Correlation)
  • NCC (Normalized Cross-Correlation)