Preserving dynamic range by advanced color histogram matching in stereo vision
Abstract
Recently, big advances have been made in developing stereo assistance systems for the production of stereoscopic movies. This paper handles the problem of possible unmatched colors between two stereo views. Based on an approach for multi-view coding efficiency, we present an advanced cumulative histogram matching method for narrow baseline stereo vision. In general results from basic cumulative histogram matching suffers from quantization errors in low dynamic range images. But also wrong colors might be introduced. To avoid such visible errors, we identify depopulated histogram ranges in color matched images. In detected ranges we adjust the mapping function by a constant value which leads to a linearized gradation curve, preserving the dynamic range of the image. Finally, we obtain an artifact free color matched image. Our approach can be applied as a pre- or a post-processing step.
