基于梯度直方图变换增强红外图像的细节

Enhancement of infrared image details based on gradient histogram transform

针对红外图像信噪比低、目标边缘和细节模糊等缺点,提出了一种基于梯度直方图变换增强红外图像细节的方法。通过分析红外图像的梯度直方图,构造出一个高斯函数来扩展梯度直方图,增大图像的梯度值。采用直方图规定化法,实现图像的梯度直方图规定化,得到变换的梯度场。在从变换的梯度场重建增强的图像时,引入全变分(TV)模型来抑制噪声。实验结果表明,提出算法的图像信息熵比灰度直方图均衡化和规定化算法显著提高,明显增强了红外图像的纹理细节,为目标检测、跟踪和识别提供了高质量的图像信息。提出的算法采用有限差分法迭代进行求解和Visiual C++编程,对大小为480pixel×480pixel的图像的处理时间约为40ms,基本达到了工程应用对图像处理的要求。

According to typical problems of low signal to noise ratio and fuzzy edges of target details in an infrared image,this article proposes a infrared image enhancement algorithm for image details based on gradient histogram transform.By analyzing the gradient histogram of the infrared image,a Gaussian function to extend the gradient histogram was constructed to increase the gradient value.By using the histogram specification method to realize the image gradient histogram specification,the transform gradient field was obtained.The Total Variation（TV）model was introduced to control the noise while reconstructing the gradient field.The experiment results indicate that the image information entropy of the algorithm is improved significantly as compared with that of histogram equalization method and specification method.Therefore,it enhances image details and gives qualified image information for target detection,tracking and identification.Based on the finite differenceiteration and Visiual C＋ ＋programming,the proposed algorithm has image processing time of about 40ms for an image with the size of 480 pixel× 480 pixel,which satisfies the requirements of engineering applications in image processing.