Improved Retinex for low illumination image enhancement of nighttime traffic

An image enhancement algorithm with improved Retinex is proposed for images captured at night that are easily affected by low light and suffer from narrow dynamic range, significant noise and loss of detail information. Firstly, the low illumination image adaptive algorithm is further proposed in combination with the traditional Retinex theory, and the global logarithmic mean is used in the calculation process, which has a very obvious adjustment effect for nighttime images. Then, the illumination part of the model is estimated using bilateral filters to obtain an accurate and reliable reflection part, so as to recover the true color and details of the image. The experimental results show that the improved algorithm is feasible and effective, and can further improve the visual effect of the image.