I am working in the field of digital image restoration. I have images that are degraded by camera shake. My images are similar to this blurry photo from camera shake. Can anyone please suggest a better non-blind deconvolution algorithm?
I have already checked some of the blind deconvolution algorithms such as Dilip Krishnan and others, but these algorithms are not producing desired result on my highly blurred images.
In our days you'd probably solve this problem using Deep Learning.
Probably with Generative assistance to mitigate the blurriness of MSE bases estimators.
On on the most significant work is in the paper Rob Fergus - Removing Camera Shake from a Single Image (MATLAB Code available).
It brought the idea of using Natural Images Statistics.
Specifically the distribution of the gradient in a natural image.
Later works just improved the priors.
Before the Deep Learning boom the priors which gave the best result were sparse inducing and low rank applied on the patch level.
See Libin Sun - Edge Based Blur Kernel Estimation Using Patch Priors.
