Abstract: Aiming at the underwater environment of nuclear power plants which leads to image quality degradation, such as colour deviation, insufficient contrast and blurred details, a method of nuclear power plant underwater image enhancement based on dominant feature fusion was proposed. On the basis of using automatic colour equalization algorithm to achieve image colour correction, the sharpness and contrast of the image were enhanced by the improved unsharpened mask algorithm and weighted adaptive Gamma correction algorithm, respectively. The advantageous feature multi-scale fusion was carried out on the sharpness-and contrast-enhanced image by using the weight map. Taking the original underwater image dataset of the nuclear power plant as a sample, a comparative experiment was conducted between this method and five other underwater image processing methods to verify the effectiveness of this method The results show that this method can effectively solve the problems of underwater colour bias, lack of contrast and blurred details in nuclear power plants, in which the mean values of underwater image quality evaluation metrics (UIQM) and information entropy are the highest in general, which are 3.103 7 and 7.502 7, respectively, and are improved by 121% and 9.66% compared with the original images.In addition, the enhanced images using the method proposed in this article significantly increase the number of feature point matching pairs, and greatly improve the efficiency of visual feature extraction and feature matching. This method can provide new technical support for underwater image analysis and intelligent detection of equipment defects in nuclear power plants.