Deblur images of a fast moving conveyor

Question

I need to process a video stream of a belt conveyor for transporting coal, mainly to detect anomalies on the conveyor for safety reasons. Because of the conveyor's fast speed, all images are very blurry. The images are like this:

I want to de-blur the images first, then run other object detection algorithms. I have tried the de-convolution script from OpenCV, the results aren't satisfying.

Any other algorithms or tools for this job?

Answer 1

Method 1

As you aim to deblur image for the sake of object detection, you can use DeblurGAN, A blind motion deblurring using conditional adversarial networks. They show how deblurring can significantly improve object detection model performance.

Method 2

Also, you can try the Photo Restoration method. It's able to remove structured and unstructured degradation effects in an image. An image like suffer from blurriness, noise, scratches, etc. Below is a quick output of your sample by this model. The right side is the raw blurry input and the left side is an enhanced version of it.