Method to Recognize Character's Color Polarity of Text Region

Characters in a text region may have different color polarities.To convert correctly the image with grayscale in an accepted text region into the OCR-ready binary image,a method is proposed to classify then recognize the color polarity of characters in a text region.The gray-gradient co-occurrence matrix of the text region is calculated,and the optimum thresholds of segmented grayscale and gradient are found quickly according to the objective function.Then,the feature vector is extracted from the gray-gradient co-occurrence matrix and fed into neural network to classify the color polarity.All the characters in the text region are finally segmented according to the classification of color polarities. Experimental results showed that the proposed method can recognize correctly different color polarities of characters in the background with different complexities.