Batch color classification using bag of colors and discriminative sparse coding

Color can be a useful feature in many fields of AI that are based on machine vision. Unfortunately, many existing vision system do not use color to its full extent, largely because color-based recognition in outdoor scene is complicated, and existing color machine vision techniques have not been shown to be effective in realistic outdoor images. The problem of color recognition in outdoor is considerable when we are faced with glossy materials like automobiles. There is no powerful method to recognize color of a batch of pixels. Thus, for the first time, we propose a novel method to detect dominant color of a group of pixels. This method has many applications in object color detection especially for glossy objects.