Activity 19: Restoration of Blurred Images

Objectives: To demonstrate the concept of motion blur and its restoration as well as basic noise filtering.

Tools: Scilab with SIP toolbox.

Procedure: We begin with an ordinary image with text written on it, in this case we use a grayscale image of the Apollo 13 Mission Patch with the words "Apollo XIII" and "Ex Luna Scientia."

We are in theory to simulate motion blur using the equation:

However, some difficulty was encountered in implementing this. As such the images were blurred using Photoshop, with varying distances of blur. After this, we add noise to the blurred image using the Gaussian image routine established in the previous activity, resulting in image like the one below:

Then using the filtering method described in the activity, Weiner filtering, the original image is reconstructed using different parameters. The working equation is given as:

The reconstructed images for the middle sample above (with medium blur) follow:

Interestingly, the filtering routine seems to have worked even if the blurring functions used are of uncertain similarity. It appears that the blurring routine used in PS is similar to the routine we are to use here, but of course the correlation may be purely coincidental.

Evaluation: For the properly reconstructed images a grade of 10 is proper.

Acknowledgements: As usual Mr. Cabello and Panganiban have rendered invaluable assistance in this activity.