No reference image quality assessments based on edge-blur measure and its applications in printed sheet blurs classification

No reference image quality assessments based on edge are important research methods, but these algorithms do not consider blur direction. This does not meet the needs of our work, printed sheet image blur classification. This paper includes two main parts. First, analyzing rules of image blurs and design a novel no reference image quality analysis based on edge, edge-blur measure (EBM). It not only takes overall-blur of image assessment, but also takes image blur assessment in some directions. Second, use these methods to classify the printed sheet blurs (sharp, defocused and shake). Extract the image blur features from EBM. Then by regression analysis, all blur features are discretized into three values. Then the discrete blur features will be input into naive Bayes classifier for blur classification. Finally, our experimental results demonstrate the effect of EBM and classification algorithm.