ADAPTIVE GAMMA PROCESSING OF THE VIDEO CAMERAS FOR THE EXPANSION OF THE DYNAMIC-RANGE

We have developed a new method for the expansion of the dynamic range of the video camera. A variable and nonlinear gamma characteristic is applied to the input image depending on the distribution of the luminance. The gamma characteristic is decided so as to amplify the luminance of the dark pixels and to preserve the contrast of the bright pixels for the back-lit objects. The output luminance equals the input luminance for the front-lit objects. We have established the decision rule of the gamma characteristic using the learning algorithms of the neural networks in order to make the decision rule coincide with the human vision. The effect of our new method is the expansion of the dynamic range by 9.6 dB. We have developed a new structure for the implementation of the gamma decision rule. It consists of a cascade connection of RAMs and its scale is less than 1/100 than the conventional structure with a single RAM. We have implemented the adaptive gamma processing in a consumer video camera.