Research on highlight restoration algorithm for sea surface oil spill images

With the advancement of digital image processing technology, the restoration of highlight images has become a significant research direction. In the case of oil spill monitoring systems for close monitoring of the sea surface, improper camera shooting positions and the presence of sunlight at a small incident angle can result in highlight areas in the captured images. These highlight regions significantly impact the quality of oil images which also reduces and the detection accuracy of oil spills. Since traditional highlight removal methods cannot be applied to water scenes, this paper proposes a Criminisi-PM highlight restoration algorithm for image enhancement. Firstly, for the improvement of the priority calculation formula, the multiplicative relation between the confidence item and the data item is changed to the addition relation. In this way, even if the confidence term is zero, the impact of data items still exists, which avoids the impact on priority when the confidence is zero, and improves the stability and accuracy of the algorithm. The introduction of the brightness term of the image in the priority calculation formula can focus more on the restoration of pixel points with larger brightness values. Secondly, for the improvement of the matching formula of the best matching block, the pixel gradient and Euclidean distance information are added to consider the features of the image more comprehensively, which makes the matching process more robust, especially in the processing of images with complex texture structures. The proposed algorithm can effectively improve the quality of visible oil spill image by eliminating the influence of the highlight region in the image. Experiments show that the method has a strong anti-interference ability and after removal the image is more natural.