Method of performing robust auto white balance

Abstract

Disclosed herein is a method of performing robust white balance. The method includes a first process of dividing an input image into blocks, a second process of selecting blocks having higher color errors, and calculating average RGB values of the selected blocks, a third process of selecting a specific top percentage of bright blocks, and calculating the average RGB values of these selected blocks, a fourth process of calculating average RGB values through color clustering, a fifth process of converting the average RGB values into CIE xyY color space, a sixth process of calculating the Euclidean distances between the x and y values of standard light sources in a color space and the x and y values of the average RGB values in CIE xyY color space, and selecting an average RGB value having the shortest Euclidean distance (Ed), and a seventh process of calculating white balance gain values using the selected average RGB value, and correcting the input image using the balance gain values.