一种基于梯度域的彩色图像转灰度图像的方法

A Color to Gray Method Based on Gradient Domain

本文提出了一种基于梯度域的彩色图像转灰度图像的方法。当彩色图像转化为灰度图像时,一些亮度相似但颜色不同的可视细节可能会丢失。本文通过在梯度域的操作,试图减少这种彩色图像转化为灰度图像时可视细节的丢失。此方法首先将图像从RGB颜色空间转化到LCrCb空间,然后对各个分量分别计算其梯度。其次对亮度L的梯度进行修改,若梯度值较大,则保持不变;若梯度值较小,并且对应位置色差分量Cr或Cb的梯度值较大,则保持亮度L梯度的方向不变,适当放大其大小。最后通过对修改后的亮度梯度求解泊松方程,得到灰度图像。实验表明,此方法获得的灰度图像保留了大部分的可视细节。

In this paper, we developed a color to gray method based on gradient domain. When a color image is converted into gray image, some visible details caused by the chromatic difference with similar luminance will be lost. We attempt to preserve these details through the processing on the gradient domain. Our method has three steps. First, it converts the input image into LCrCb color space and calculates the gradients of every channel. Secondly, the small gradients of L channel are augmented if the corresponding gradients of Cr or Cb channel are large, and the others keep unchanged. Finally, the gray image is obtained from the modified gradients of L channel by solving a Poisson equation. The experiments show that the most visible details are preserved in the gray image.