Colour constancy using sub-blocks of the image

Abstract

Colour constancy is the ability to measure the colour of objects independent of the light source, while colour casting is the presence of unwanted colour in digital images. Colour casting significantly affects the performance of image processing algorithms such as image segmentation and object recognition. The presence of large uniform background within the image considerably deteriorates the performance of many state of the art colour constancy algorithms. This paper presents a colour constancy method using the sub-blocks of the image to alleviate the effect of large uniform colour area of the scene. The proposed method divides the input image into a number of non-overlapping blocks, and Average Absolute Difference (AAD) value of each block colour component is calculated. The blocks with AAD greater than threshold values, which are empirically determined for each colour component, are considered to have sufficient colour information. The selected blocks are then used to determine the scaling factors to achieve achromatic values for the input image colour components. Comparing the performance of the proposed technique with the state of the art methods using images from three datasets shows that the proposed method outperforms the state of the art techniques in the presence of large uniform colour patches.