直方图均衡类算法由于缺乏限制项, 增强后的红外图像易出现灰度级合并和过增强, 视觉效果不佳。为解决上述问题, 提出基于场景复杂度限制型的平台直方图均衡算法。该算法首先提出对图像场景复杂度进行估计, 并将Harris角点数量作为图像复杂度评价依据。然后建立类似sigmoid函数的图像复杂度评价函数, 将场景复杂度标准化。最后综合考虑图像复杂度和防止原图均值漂移, 依据平台直方图均衡原理获取全局变换函数, 重建红外图像。实验结果表明: 该方法可有效地依据不同场景实现红外图像增强, 不出现过增强和噪声放大。客观评价指标表明, 该方法在保持图像均值亮度和图像信息熵方面表现优异。

Histogram equalization and its improved version may cause tinny gray levels coalition and over-enhancement in the output infrared image without any constraints. In order to solve this problem, we propose plateau histogram equalization based on scene complexity computation. Firstly, Harris algorithm is used to detect corner. Then an evaluation function, which is similar to the sigmoid function, is established to calculate the degree of complexity for an image. Finally, a global transformation function is gained using the theory of plateau histogram equalization. Moreover, the global transformation function achieves enhancement with the constraints of image complexity and mean lighting protecting. Experimental results show that the proposed method can enhance infrared images with different scenes effectively and there exist no over-enhancement and noise amplification in the output infrared images. Objective evaluation illustrates that the proposed method can achieve less mean lighting shift and holding the entropy of original image.