Abstract

PCA-BASED BLIND RESTORATION OF FACE IMAGES USING GRADIENT PROFILE PRIOR MODEL

This paper presents a face image restoration algorithm by using gradient profile prior model. Basic blind image restoration scheme applied here is our PCA-based method where further blurred images from a given degraded image are analyzed and then clear image is estimated by optimizing image quality index. Our previous method is modified for targeting face images and utilizing the gradient profile prior. For this purpose we first reveal some generic gradient profile features observed in face images. Through these investigations we introduce a novel statistical model for the prior. The final restoration procedure succeeded by applying PCA is to determine the weights of the principal components by maximally matching them to face image gradient profile prior. Some experiments are conducted to demonstrate the effectiveness of the proposed algorithm.

Keywords— Blind image restoration, Face images, Image gradient distribution, Motion blur, PCA subspace