根据感受野机制,人眼对场景中各点的感知程度是呈同心圆拮抗式的高斯分布。受此启发,本文提出一种基于人眼视觉特性的Retinex算法,即同心圆拮抗式Retinex算法。该算法在以待处理点为中心的同心圆区域使用高斯分布对像素点进行采样,计算样本中的最亮点,并将该点与处理点进行亮度比较,计算出物体的反射分量。该算法有效解决了基于路径Retinex算法中,对路径强烈依赖和无法获取尽可能多邻域信息等问题,避免了中心环绕Retinex算法中存在的全局性较差的问题。实验结果表明,本文算法在全局和局部上,都有很好的增强效果和色彩保真性,同时也避免了"光晕伪影"。
According to receptive field mechanism,the perceived degree of the human eye to the points in scene presents the concentric circles with antagonism Gaussian distribution.Inspired by that,a new Retinex algorithm based on human visual property is proposed in this paper.Retinex algorithm can also call concentric antagonistic type Retinex algorithm.We sample the pixels in the concentric circles area which take pending point as the centre with the Gaussian distribution.Calculating the highlight of the samples,then the difference of lightness with the pending point to restore reflective component of the objects is calculated.The proposed algorithm can solve the problem that Retinex algorithm based on path strongly depends on the path and can't get as much information from the neighborhood pixels.Moreover,the problem that Centre/Surround Retinex has got bad global properties can also be escaped by this algorithm.The experimental results show that the proposed algorithm has wonderful enhancement effect and color fidelity in the globle and local properties.Moreover," halo artifacts" can be also escaped.