Image Mosaic Based on Local Feature and RANSAC

This project is an implementation of image mosaic based on local feature descriptors and RANSAC, coded in Python language.

Image mosaic is to make a panorama from given images.

Project Structure

• Runnable Python source file is ImageMosaic.py, implemented with SIFT/SURF/ORB and RANSAC algorithm. JUST CLONE THE REPOSITORY AND RUN IT!
• Images to be stitched are in dataset directory, which contains 2 types, one is translation, another is rotate-scale.
• The output directory contains panorama results generated by ImageMosaic.py. Images in the unblend directory are the mosaic results without seam elimination.

I. Task of Image Mosaic

II. Principle of Image Mosaic

III. Process of Image Mosaic

IV. Tips for Image Mosaic

V. Seam Elimination and Optimization

VI. Implementation

VII. Results

• Result of Translation

• Result of Rotate-Scale

VIII. Crop to Make More Standard Panorama

Dependency

• OpenCV 3.4.1

References

• [1] Brown, M., Lowe, D.G. Automatic Panoramic Image Stitching using Invariant Features. Int J Comput Vision 74, 59¨C73 (2007). https://doi.org/10.1007/s11263-006-0002-3
• [2] David Kriegman. Homography Estimation. Computer Vision I. CSE 252A, Winter 2007
• [3] https://github.com/kushalvyas/Python-Multiple-Image-Stitching

Author Info

LeoHao (XMU-CS)

Date

2020.12.10