Chang, C.-F., Wu, J.-L., and Tsai, T.-Y. A single image deblurring algorithm for nonuniform motion blur using uniform defocus map estimation. 2017
Not finished implementation.
Install dependencies:
pip -m install requirements.txt
Run program:
cd MotionBlur
python3.10 main [--input filepath] [--output filepath]
The implementation is executed in three main steps.
The first step is to compute the boundary points of the colour image in the CIE Lab colour space. The boundary points are computed using the Sobel operator. After applying horizontal and vertical operators, the gradient image is obtained. The means and variances of this image are computed, and by averaging the values, the new final gradient image is obtained. In the paper, the threshold function wasn’t specified, so I applied thresholding based on the percentile.
In the second step, the grayscale images are created. They are created by using linear operations and ratios between red, green, and blue components. The step of change for the components is set to 0.1, which means there are 66 combinations of the components in total. The boundary points are computed from each grayscale image as well.
The final step consists of comparing boundary points from a colour image and boundary points from grayscale images. The selected grayscale image is the one with which the colour image has the same boundary points (most of them). And finally, the image is stretched to fit the correct range
Install dependencies:
pip -m install requirements.txt
Run program:
cd ColorToGray
python3.10 main [--input filepath] [--output filepath]
