Implementing exposure control by maximizing a robust gradient-based image quality metrics to improve robustness of visual Odometry in challenging illumination conditions using HDR images. This project was carried out as a group project in a team of 2.
This repo contains source codes and a report with detailed explaination of the project and code implementation.
- The program uses OpenCV and is written in C++. It implements the control algorithm by optimising the image quality metrics.
- The program uses a hardcoded datase. The optimised exposure time for the next frame is calculated based on the dataset and is not augmented.
- The program calculates two different gradient based image quality metrics and verifies the best frame by extracting fast features.
- The optimisation algorithm is the gradient-ascent method.
- The photometric response calculation needs a minimum of 10 frames failing which the response will not be calculated or will be inaccurate.
- A change in image format must be altered in the code as the dataset is hardcoded. The program only accepts images of the resolution 512x768 pixels.
- mkdir build
- cd build
- cmake ..
- make