Pipeline Image Dehazing Algorithm Based on Atmospheric Scattering Model and Multi-Scale Retinex Strategy

Abstract

At present, the detection video of the drainage pipeline is often affected by the water mist in the pipeline. As a result, a large number of pipeline detection videos are highly blurred, which seriously affects the detection effect. In this paper, we proposed a algorithm based on atmospheric scattering model and Multi-Scale Retinex strategy. Firstly, the threshold is used to select the atmospheric light in the fuzzy image. The algorithm then uses the estimated depth information to calculate the transmission map, which is substituted into the atmospheric scattering model. To overcome the issue of low image contrast, we improved the calculation of transmission by adding loss compensation factor. In addition, we use Multi-Scale Retinex strategy as color restoration algorithm to enhance the improved dark primary color method to further improve the image clarity. The experimental results show that the algorithm has good effect on the image of underground pipeline with large reflection area. In the future, the algorithm will be applied to pipeline robots to ensure the normal operation of underground pipelines.