基于Retinex算法的中波红外图像区域分割

Region segmentation of intermediate wave infrared image based on retinex algorithm

中波红外图像的采集过程中,由于受地球固有特性以及采集仪器自身的特性等因素影响,导致图像含噪较多,易出现观测失准现象。为提高中波红外图像质量,需要对其进行分割处理,传统方法往往忽略了图像的噪声特征,未考虑对噪声的处理,导致其分割精度较差。提出基于Retinex算法的中波红外图像区域分割方法,首先提取中波红外图像的噪声特征,确定最小噪声量,利用Retinex算法对图像噪声特征进行增强处理,当输出增强噪声小于最小噪声量时,即实现去噪处理。通过自适应高阶特征匹配,利用高阶像素特征,实现对中波红外图像的分割,达到准确高效分割中波红外图像的目的。实验结果表明,采用所提方法对中波红外图像进行分割处理时,其抗噪能力较强,耗时短,且均方根误差低于传统方法,具有更高的分割性能,可有效改善中波红外图像质量。

Due to the influence of the inherent characteristics of the earth and the characteristics of the acquisition instrument, the image contains more noise and is prone to misalignment in the acquisition process of the medium-wave infrared image. In order to improve the quality of mid-wave infrared image, it is necessary to segment it. The traditional methods often ignore the noise feature of the image, and do not consider the noise processing, which leads to the poor segmentation accuracy. A region segmentation method of middle wave infrared image based on Retinex algorithm is proposed. Firstly, the noise feature of medium wave infrared image is extracted, the minimum noise is determined, and the noise feature of image is enhanced by Retinex algorithm, and the output is increased. The experimental results show that the method can effectively improve the quality of the medium-wave infrared image by using the proposed method to divide the medium-wave infrared image. The experimental results show that the method has the advantages of stronger anti-noise ability, short time consumption and lower root mean square error than that of the traditional method , and can effectively improve the quality of the medium wave infrared image.