Image blurring and deblurring using simplicity of spatial domain over time domain
This video enables one to understand how kernel blur filters can be created if you know the standard equations of different function, like I know the standard equations of Gaussian and laplacian of Gaussian. So I can code to create blur or one can also call it point spread spectrum (PSF). Matlab image processing toolbox is an effective tool to understand operations in computer vision in better way. I have made use of concept that Convolutional in time domain is equal to multiplication in frequency domain and hence fast Fourier transform of image and created kernel is implemented and in the last to convert back the result in the time domain, inverse fast Fourier is calculated. For deblurring the blurred image is divided by the same kernel which has deformed it. In this code, user has two choices for kernel which you can see during video. I have used the equation of both the Gaussian and laplacian of the Gaussian to construct the kernels/PSF to blur the image. I will show the demo in the video.
Now let us see the coding part. So different choice will produce different results. You could see in the code. Now please focus on the next image
Sometime first method is good and sometime second method is good. Please see the equations of both the kernel function.
Hope you like the video. Thanks from AI Sangam.