一种基于误差测度的累进颜色量化方法

摘要

提出一种新的彩色图像颜色量化方法。该方法基于全局最小误差测度实现图像中的颜色数递减，直到满足规定的颜色数。图像最终可量化为一种颜色，该颜色是所有图像颜色的加权平均。误差测度考虑所有可能的颜色合并所产生的误差，并选择误差最小的颜色对进行合并。合并后产生一种新的颜色，可从理论上证明这种合并方法所产生的图像其颜色值畸变最小。实验验证了该方法的有效性，产生了很好的视觉效果，在真彩图像分割、目标识别等领域具有研究意义。%We propose a new colour quantisation method for colour images.This method achieves the degression of colour numbers based global minimum error metrics until the specified colour numbers are satisfied.The image can be quantified as one colour eventually which is the weighted average of all colours in image.Error metrics considers the errors caused by merging all possible colours,thus those colours with minimum error are selected to merge.A new colour will be produced after merging and it is able to prove theoretically that the image generated with such merging method has minimum distortion in colour value.Experiments verify the effectiveness of the method,a very good visual effect is realised.This method has the significance of research in many fields,such as true-colour image segmentation and object recognition.