Content and Illumination Invariant Blur Measures for Realtime Video Processing

This paper presents an approach to optical blur estimation in images based on the measuring of the spread of edges. Two measures based on this approach are proposed. The first measure defines a model of amplified Gaussian with background for the profile of edges. The second measure is an approximation that does not require the extraction of profiles at edges. Both measures do not depend on the content or illumination of the image, making them suitable for dynamic videos. The behaviour of the proposed measures is finally presented in the context of dynamic videos from a surveillance camera embedded in a transportation vehicle.

[1]  Michel Desvignes,et al.  Mesure de netteté par Transformée en ondelettes. Définition et comparaison pour l'autofocus de caméra , 2003 .

[2]  Steven W. Zucker,et al.  Local Scale Control for Edge Detection and Blur Estimation , 1996, IEEE Trans. Pattern Anal. Mach. Intell..

[3]  Stefan Winkler,et al.  Perceptual blur and ringing metrics: application to JPEG2000 , 2004, Signal Process. Image Commun..

[4]  Wilfried Philips,et al.  Estimating image blur in the wavelet domain. , 2001 .

[5]  Sei-Wang Chen,et al.  A non-parametric blur measure based on edge analysis for image processing applications , 2004, IEEE Conference on Cybernetics and Intelligent Systems, 2004..

[6]  Frédéric Guichard,et al.  Uniqueness of blur measure , 2004, 2004 International Conference on Image Processing, 2004. ICIP '04..