An automated defect management for document images

This paper describes a new approach for automated quality improvement of grey-scale document images, called STORM. The grey-scale images are first adaptively partitioned into representative regions, whose content is analyzed using a set of document image features developed and adapted for the purpose. The document condition and quality information is evaluated for defect pattern classification in a given entity. This data is then processed using a neural network classifier to expose and prioritize the image defects, if any. The evaluation information is further partitioned using the soft control technique by mapping and parametrising the evaluation classes into available image operation techniques. The document type and domain characteristics are used to bias these operations. The experiments cover over 1000 document images in different categories having degradation types in various degree. The outcome shows good results in most of these domains with an automated process