I am faced with the challenge of both reducing glare from headlights and increasing overall contrast for better object recognition in an attempt to enhance the image quality of road vehicles at night. I am brainstorming here so I have not tried anything yet but plan on using OpenCV and Python as well as making hardware/camera choices to suit the application.
Hardware as an important aside, my question here is in regards to image processing techniques for low latency video streaming
** I am not attempting to classify objects using bounding boxes, I am attempting to clean up the images for humans to view. This project is for an ADAS system using a large rectangular screen displaying the rear view form the vehicle for better night driving experience. Thermal Cameras are also currently outside the scope of this project
Histogram Equalization is said to help extract data from images with poor contrast ( Cars on a dark road ) but what about reducing the blinding effects of headlights?
HSV + CLAHE seems to be a common place to start but no universal method seems to be in use to do the innate complexities of the problem.
What am I getting wrong?