Improved multi-scale Retinex infrared image enhancement

In order to solve the problems of halo phenomenon and image noise enhancement when the traditional multi-scale Retinex algorithm enhances the details of infrared images, an infrared image detail enhancement algorithm based on improved mult-iscale Retinex is proposed. Firstly, an improved guided filter is used instead of the Gaussian filter to accurately estimate the incident component of the infrared image and calculate its reflection component to reduce the noise in the image and improve the halo phenomenon. Secondly, an adaptive grayscale stretching process is performed on the incident component, and a smoother non-linear curve is used to correct the reflected component, so as to achieve the purpose of better enhancing the overall brightness and contrast of the infrared image. The experimental results show that compared with other corresponding algorithms, the information entropy, peak signal-to-noise ratio, average gradient, and standard deviation have been improved. At the same time, the infrared enhanced image obtained by this algorithm has higher contrast and clearer details. The human eyes have better visual effects.