An efficient and effective background determination for color document images

Conventional global thresholding methods have three shortcomings. First, most of them are not efficient for processing a large number of color document images. Second, most of them assumed the gray level of the background is higher than the foreground. In real color document images, the gray level of the background is lower than the foreground. Furthermore, some color document images, the backgrounds are in the middle of the gray level. Third, most of them do not consider the spatial characteristics. Thus, this paper proposes an efficient and effective background determination method for color document images to solve aforementioned problems. The proposal method includes following steps: First, RGB color space is transformed to gray level. Second, the boundary between the background and the foreground is determined by scan line differentiation algorithm. Third, the background and the foreground distributions in the ambit are analyzed to determine the background ranges. Finally, the document image is converted into binary image by the background ranges. This novel method is very efficient because it only analysis the border pixels in the color document image. Our experiments include various color document images. Experiment results show that the performance of the proposal method is better than other available methods in visual measurement.