Motion blur detection in static images.

摘要

Digital photography has become very popular and supersedes the traditional film-based photography. Automatic quality assessment of digital images taken by digital cameras or camera phones is consequently desirable, since ordinary users are hardly capable of determining the clearness via small screens installed in these devices.; Previous research mostly focused on the estimation of blur parameters. This thesis is an attempt to propose an effective scheme that can decide whether an image is blurry or not. The classification process is based on characteristics of feature point orientation histograms. We investigate several feature point extraction approaches, and use wavelet transform coefficients to compute the orientation of each feature and its neighbors. Support Vector Machine (SVM) is then applied to construct classifiers for orientation histograms. Experimental results are given and the performance discussed for both high resolution images taken by good digital cameras and low resolution images taken by camera phones.