Color-distorted image dataset

Automatic white balance (AWB) is an important module for color constancy of cameras. The classification of the normal image and the color-distorted image is critical to realize intelligent AWB. One tenth of ImageNet is utilized as the normal image dataset for training, validating and testing. The distorted dataset is constructed by the proposed theory for generation of color distortion. To generate various distorted color, histogram shifting and matching are proposed to randomly adjust the histogram position or shape. Furthermore, the extent for shifting or matching is randomly generated to ensure the diversity of color distortion. The dataset is robust because both histogram position and histogram shape are randomly adjusted. There are totally 288,132 images in the dataset.