MULTI-FACETED IMAGE RESTORATION AND ENHANCEMENT

EE 610: Image Processing, Prof. Amit Sethi [Github]

1. Image Denoising using Neighbouring Wavelet Coefficients

Utilized the approach of Chen-et-al Image Denoising using Neighbouring Wavelet Coefficients to effectively for practical image denoising applications. This approach utilizes the correlations between neighboring wavelet coefficients, i.e., a large wavelet coefficient will probably have large wavelet coefficients as its neighbour.

2. Motion Deblur

Predicted and implemented the motion blur kernel by utilizing the angle of smear present in motion-blurred car images to give the direction of motion blur. Finally, performed Weiner filtering using the estimated blurring kernel to effectively reveal the license plate characters in these shaky images of cars.

3. Satellite Image Restoration

Performed satellite image restoration by removing subtle scanning artifacts of one of the images taken by Voyager 2 by filtering the images piecewise in the Fourier domain. Revealed intricate details by histogram manipulation while simultaneously maintaining color balance.

4. Moiré Effect

Performed filtering in the Fourier domain to eliminate the moiré effect. The 2-D Fourier transform of the image had high values present on the X-axis and Y-axis due to high frequency vertical and horizontal components in the input image. Masking these values resulted in the complete removal of the moiré effect.