A New Structural No-Reference Rule Based Blur Metric for Classification of Blurred Home Photos

Abstract

Automatic classi¯cation of blurred photos from family album has drawn researchers' attention to avoid computational overhead of deblurring and measurement of blur in the photos. Therefore, the primary goal of this paper is to classify the blurred photos rather than measuring degree of blurness in the image or to deblur an image. The blur metric is proposed based on the fact that the number of Canny edge components is more than the number of Sobel edge components in blurred photos. Similarly, for non blurred photos, the number of Canny edge components is less than the number of Sobel edge components. The rules are devised based on the width of Canny and Sobel edge components to classify the blurred photos. The experimental results of the proposed metric are compared with the results of the existing methods for the purpose of evaluation.