中文摘要：

颜色恒常性计算就是通过消除光照对颜色的影响，得到与光照无关的稳定的颜色描述因子。目前基于底层特征驱动的颜色恒常性算法，大多数利用整幅图像的像素信息来估计场景的光照。然而，并不是所有的像素点都包含颜色恒常性计算的有效信息，而且没有考虑像素点周围的空间信息的影响。本文针对这两个问题提出一种基于最优区域选择和离散小波变换结合的颜色恒常性算法，在分割区域上进行不同尺度的离散小波变换，利用得到的小波系数估计出不同区域不同尺度上的误差，选择出误差最小的分割区域用于整幅图像的光照估计。该算法简单易行，实验结果证明可以取得比较好的光照估计效果。

英文摘要：

The purpose of color constancy computations is to eliminate the illumination influence on colors, and to get the stable light-independent color descriptors. Most color constancy algorithms based on the low-level features, estimated the illumination in the scene by the whole image pixel information. However, there are two problems： 1） not all the pixels contain the effective information for the color constancy computation; 2） these methods did not consider the influence of the spatial information of the surrounding pixels. Due to above problems, an algorithm based on the optimal region and the discrete wavelet transform is proposed. Firstly, the discrete wavelet transform is performed on the segmentation regions of the input image at each scale. Secondly, the il-luminant errors are estimated by the wavelet coefficients at each scale. Lastly, the whole image illuminant color is estimated by choosing the segmented regions with the least errors. The proposed algorithm is feasible. The experimental results prove the good illuminant estimation for the proposed method.