Day markings detection in natural photo with morphological method

摘要

Day marking is one of valuable information for many applications such as image retrieval and indexing. In this paper, a novel automatic scheme is presented for day marking detection from color photographs. First, taking advantages of the stroke thickness, gray-scale morphologic method is specially proposed to generate the bounding box in R-component domain, where the day texts are located and the background influence can be reduced. Then binarization module is set to favor character recognition, where some post-processing is taken to remove white-dot noise and fill the holes in the strokes. Finally the binarized image should be segmented and fed to OCR module so that the characters of the day marking can be recognized. Simulations show the effectiveness of our method.