基于先验知识的单幅图像天空检测算法

A Single Image Sky Detection Algorithm Based on Prior Knowledge

为了对图像中的天空区域进行有效识别,提出像素偏转模型评价三通道像素灰度值的特点,并结合天空先验知识进行天空区域识别。像素偏转模型可通过分析RGB三通道像素点灰度值特点,简单、准确地评价像素点3个通道之间的离散程度。天空先验是根据对大量图片中天空部分像素点的统计得出的,通过对图片中天空部分的观察发现,将天空先验与像素偏转模型相结合,可以直接判断出像素点是否为天空区域,从而准确识别出图片中的天空。该算法具有较强的鲁棒性,不仅能够识别边界清楚的天空区域,而且对于边界复杂的天空区域也有较好的识别效果。

In order to effectively identify the sky area in the image.In this paper,a pixel deflection model is proposed to evaluate the characteristics of the gray values of the three channels,and the sky region is identified by combining the sky prior knowledge.The pixel deflection model can simply and accurately evaluate the dispersion degree between the three channels of the pixel points by analyzing the gray values of the pixel points of the RGB three channels.Sky priori is obtained from the statistics of the sky part in a large number of images.Through the observation of the sky part in the image,we find that combining this priori information with the pixel deflection model,we can directly judge whether the pixel point is a sky region and accurately identify the sky in the image.The algorithm has strong robustness,which can not only identify the sky region with clear boundary,but also has good recognition effect for the sky region with complex boundary.