多尺度暗通道先验去雾算法

目的提出了融合块部重叠区域生长的多尺度暗通道先验去雾算法。方法为了改善暗通道先验去雾算法的计算量，将多尺度的思想引入暗通道先验去雾算法。由于天空区域不满足暗通道先验的假设，导致天空区域出现色彩失真。提出块部分重叠的区域生长算法分割天空区域，根据天空区域透射率而设定全局透射率的下限。结果实现单幅图像的去雾。结论实验结果表明，改进算法提高了去雾算法的处理速度，减少了资源的消耗，而且在天空区域的处理上，改进算法结果视觉效果更好。

Image dehazeing using multiscale dark channel prior

Objective To present an algorithm which merges together partially overlapped sub-block region growing and multiscale dark channel prior for image dehazing. Methods In order to improve the computation of dark channel prior dehazing, the duced. Because sky area does not meet the dark idea of muhiscale dark channel prior dehazing was introchannel prior assumption, the color of sky area is distorted. Therefore, partially overlapped sub-block region growing algorithm was used to segment the sky area, and the lower limit of global transmission value was set up according to the transmission value of the sky area. Results Single image dehazing was realized. Conclusions The experiment demonstrates that the speed of dehazing is increased, the resource consumption is reduced, and the sky area has better visual effect.