Comparison of different image processing techniques for VIS-NIR images in maritime environments at night

Abstract

In the maritime environment, visual (VIS) and infrared (IR) imaging systems are used for various applications, including navigation, situational awareness, and search and rescue operations. These scenarios share the need for object detection, especially in harsh environments at night and bad weather conditions. This study presents a comprehensive comparison of post-processing techniques, such as sensor specific noise reduction by a non-uniformity correction (NUC), stochastic (random) noise reduction, with Gaussian and fast Fourier transform (FFT) filtering, and contrast enhancement, with three tone mapping operators (Reinhard, Mantiuk and Drago). The experimental evaluation comprises analysis of image quality metrics such as signal-to-noise ratio (SNR) and contrast evaluation. As a result, a decision must be made as to whether reduce the noise or increase the contrast, as the combinations of the methods presented cannot realize both improvements at the same time. Overall, image post-processing can improve the detection of objects at sea and night for a human operator.