基于三直方图均衡的SF_(6)红外图像对比度增强方法

Contrast Enhancement Method of SF_(6)Infrared Image Based on Tri-histogram Equalization Algorithm

电力设备SF_(6)泄漏气体红外图像容易淹没在背景中,肉眼难以辨识低对比度图像中泄漏气体情况,给电力设备检修带来了困扰。提出了基于三直方图均衡的SF_(6)红外图像对比度增强方法。首先,利用三样条插值拟合图像直方图得到二阶连续曲线,计算每个灰度级对应的一阶导数绝对值。其次,根据一阶导数绝对值和直方图分布划分直方图为两个波峰和一个相对平坦的波谷区域;最后,根据波峰和波谷将直方图分为3个子图,对3个子图分别进行直方图均衡后合并为增强图像。为了验证算法有效性,对现场拍摄的电力设备SF_(6)泄漏低对比度红外视频流进行增强,并与CLAHE和双直方图均衡算法进行对比。实验结果表明:本文方法在提升图像的整体对比度同时增强了泄漏气体与周围图像的对比度,提升泄漏气体红外图像视觉效果,相比于CLAHE和双直方图均衡方法,本文获得的峰值信噪比和均方根对比度值更高,增强后的图像质量更好。

Infrared images of sulfur fluoride(SF_(6))leaked gas in power equipment are easily merged with the background,and it is difficult for humans to identify the leaked gas in low-contrast images,rendering it difficult to maintain power equipment.A low-contrast enhancement method for SF_(6)leaked infrared image based on tri-histogram equalization is proposed.First,a cubic spline interpolation is used to fit the image histogram to construct a second-order continuous curve,and the absolute value of the first derivative of the curve corresponding to each gray level is calculated.Second,the two extreme points of the histogram are calculated according to the absolute value distribution of the first derivative and histogram.The extreme points divided the histogram into two peaks and relatively flat valley regions.Finally,the image is divided into three sub-images according to the tri-histogram,and the three sub-images are histogram-equalized.The sub-images are merged into a new image,which is the enhanced image.To verify the validity of our algorithm,an infrared video of power equipment with SF_(6)leakage,which is recorded in substation fields,is tested to enhance the contrast.Our algorithm is compared with the CLAHE and bi-histogram equalization algorithms.The experimental results show that our method enhances not only the globe contrast for infrared images but also the local contrast for leaking gas.Therefore,the infrared image of the SF_(6)visual effect is improved.Compared with CLAHE and the bi-histogram equalization method,our method obtained a higher peak signal-to-noise ratio and root-mean-square contrast values,and the enhanced image quality of our method was better than that of the other methods.