Classified Image Enhancement Method Based on Histogram Characteristics in YCbCr Color Space

Lijing Tong¹, Jingzhong Wang¹, Sam Li², Ke Xiao¹, Quanyao Peng³

COMPUTER MODELLING & NEW TECHNOLOGIES 2015 19(5A) 7-13

¹College of Information Engineering, North China University of Technology, Beijing, China
²Palo Alto Networks Incorporation, Santa Clara, California, USA
³Beijing 58 Information Technology Co., Ltd, Beijing, China

Color image enhancement in YCbCr space is an important task since most of the color image signals captured from the embedded camera or the professional video device are YCbCr image signals. Prior classical color image enhancement methods like linear transforms such as binarization, piecewise-line transform, and gray-level slicing, or non-linear transforms such as logarithm transform, index transform, and power-law transform did not consider possible histogram characteristics, and thus their enhancement performance on different image types would be degraded in some cases. In this paper, a novel classified image enhancement method based on CbCr and Y histograms is proposed to address the aforementioned problem. First, captured images are divided into two types, document image and scene image, according to the normalized chrominance histogram characteristic. For the document image, a filter is applied in space domain to get a better foreground and background. For the scene image, three different types are divided by the normalized luminance histogram characteristics. Then, three different processing schemes are applied to the three types of scene images respectively. Experimental results on different images with a variety of variations verify the effectiveness and robustness of the proposed method.