An improved color cast feature and feature extraction method based on lab chromaticity histogram

Color is one of the most important cues for image content analysis. However, there is often a color cast between the real color of objects and the acquired color by imaging device. In order to detect color cast accurately, the color cast feature base on Lab chromaticity histogram is improved in this paper. On the basis of classic equivalent circle (EC) feature and NNO (Near Neutral Objects) region feature, the distribution characteristic in the height direction of chromaticity histogram is further considered, as well as the feature change rule between the original image and the NNO image. Then, the improved color cast feature extraction algorithm is presented. The experimental result shows that the effectiveness of the proposed color cast feature in color cast evaluation.

I. Introduction

When observing the world by human beings, the visual neural response to color, shape, texture and detail is gradually decreased. Moreover, compared with texts and sounds, color images have more rich content and more powerful expression capability. Thus, color is one of the most important cues to obtain information of the objective world. However, with the changes in ambient lighting, intrinsic reflection characteristic of the object itself, or photosensitive coefficient of imaging sensor, the acquired color by imaging device will deviate from the real color of objects, i.e. color cast [1], [2]. The color cast will degrade the visual appearance of image, and the follow-up image processing such as image segmentation and target recognition. Hence, color cast detection and color correction are usually indispensable step in image processing and computer vision.