Progress Report I

November 14, 1997

Basically, we have started to read up on the different deconvolution filters in a detail and also started to experiment with the different tools that matlab has to offer. We have decided to only restore images in greyscale as colors complicate the problem significantly. The three aspects that we will focus on are linear motion correction, deblurring and noise reduction. We plan to concentrate our efforts on deblurring in cases where the original blurring filter is unknown.

Specifically, we have looked at the Wiener filter and how to identify the system using the point spread function(PSF). As we are looking at blind deconvolution, identification of the system is fundamental to restoration of the image. Multiple versions of the same image might also help us to make better estimates and improve the blind deconvolution process.

A possible application of deblurring filters to an image with unknown properties might be an automatic system to determine the optimal filter. If a basic deblurring filter is used, paramaters such as an initial guess and a range of operation can be defined that the filter could be run via a computer program at different values until the 'best' image is found. To automate the analysis process, edge detection would be used to find this 'best' output image, such that the optimal filter with the largest prescence of edges would be found in the image with the best focus.

Word.