Restoration of blurred objects using cues from the alpha matte
Contents
Part 1: project profile
Project name
Restoration of blurred objects using cues from the alpha matte
Project short description
The project aims at the restoration of digital images containing blurred objects. We use natural image matting techniques to separate object and background data, and take advantage of the characteristics of the computed alpha matte to provide an estimation of the blur PSF (Point Spread Function). The image of the blurred object is then restored using the estimated PSF. We validated the proposed technique with both synthetic and real camera images.
Dates
- Start date: 2007-05-01
- End date: 2008-04-24
People involved
Project advisors
- Giacomo Boracchi
- Vincenzo Caglioti
- Alessandro Giusti
Students who worked on the project in the past
Part 2: project description
We propose a technique for estimating the blur point spread function exploiting the information provided by the alpha matte, when a single image containing a blurred object is available. The alpha matte contains information about the apparent transparency of the blurred object. The main idea consists in estimating the unknown silhouette of the blurred object by thresholding the alpha matte and obtaining then the blur point spread function with the deconvolution between the blurred and the thresholded alpha map. Our approach has been validated with experiments on both synthetic and camera images. On synthetic images, the blur point spread function has been estimated with reasonable accuracy even in presence of additive white noise, while on camera images, the restoration performance using the estimated point spread function encourages further works in this direction.
