mosa

[WIP] Python Image Processing module to stitch a series of similar images to create a mosaic.

Dependencies

• OpenCV 2.4
• Python 2.7

Demo

To run an interactive demo. Make sure you have OpenCV 2.4 with python bindings loaded in PYTHONPATH. There is a script that'll walk you through the stitching process in a GUI.

cd to/mosa/path
python scripts/demo.py

If you have the dependencies installed, this should pop up the hummingbird image with the features found and linked. To go to the next frame, type q.

Benchmarking

You can benchmark different OpenCV feature detection and matching algorithms to see how they perform against each other given two images.

The benchmarking script will output a CSV file and mix and match the algorithms specified in the benchmark.yml file in here. The result are the of benchmarking on two images are:

• Number of keyponints
• Time to find keypoints
• Number of features
• Time to find features
• Number of matches
• Time to find matches
• Number of outliers
• Number of KNN matches
• Time to find KNN matches
• Number of KNN outliers

Right now, only the following algorithms are supported.

• Detectors/Descriptors
  • SIFT
  • SURF
  • FAST
• Matchers
  • FLANN

To run the benchmark, do the following:

cd to/mosa/path
python scripts/benchmark/benchmark.py path/to/image1 path/to/image2

The above script will generate a out.csv file in most/scripts/benchmark.