/image-stitching

Panoramic image stitching program

Primary LanguagePython

Panoramic Image Stitching Program

The pipeline is composed of 4 parts:

  1. Feature detection - Harris corner detector
  2. Feature description - Simple custom descriptors or SIFT
  3. Extimation of the affine transformation matrix - RANSAC algorithm
  4. Transforming and overlaying the second image over the first

Qualitative Results

alt text alt text alt text alt text

Structure and Instructions

Directories:

  • Test cases: Contains pairs of input images for testing
  • Results: Contains plots and logs from all experiments

Code:

  • main.py: The main file for the application
  • utils.py: Collection of utility functions
  • descriptors.py: Implementation of feature descriptors
  • ransac.py: RANSAC implementation
  • visualizer.py: Visualization code

Instructions:

  1. To use one of the supplied test examples in base case configuration, execute:

$ python3 main.py --img1 ./test_images/Original/1.jpeg --img2 ./test_images/Original/2.jpeg

  1. To use the hard test example, execute:

$ python3 main.py --img1 ./test_images/Hard/1.jpeg --img2 ./test_images/Hard/2.jpeg --harris_thr 0.01 --descriptor "custom_rgb_intensities" --patch_size 21 --matching_threshold 0.940 --ransac_sample_size 3 --ransac_n_iterations 1000 --ransac_tolerance 50 --ransac_inlier_threshold 5