Selective color adjustment using a neural network as a color classifier

One of the subtle problems in color image processing is the difficulty of adjusting specific color tones in the image without affecting all other colors. The usual method is to adjust the primaries so as to produce as realistic as possible skin tones, and at the same time balance the rest of the colors close to their natural appearances. The reality, however, is that a balance of all color tones to their natural appearances by just adjusting the three primaries is not always a possible task. An alternative approach is to selectively adjust specific color tones without altering the color balance in the rest of the image. Such a scheme can be referred to as selective color tone adjustment. It is also possible to selectively sharpen or blur a specific color by selective color filtering. In this paper we present a scheme for color classification and apply it to selective color tone adjustment