Research On Underwater Image Processing Technology In Complex Environments

Underwater natural resources not only have great economic value but also have very important strategic significance.Underwater optical images are widely used in underwater exploration because of their rich details.The underwater environments in natural scenes are very complicated.A large number of suspended particles in the water have a strong absorption and scattering effect on light.In underwater environments,the attenuation rate of light with different wavelengths is also different.Therefore,underwater images usually have the characteristics of low contrast and blue-green color shift.To improve the perception ability of underwater optical imaging systems,some intensive studies of underwater image processing technology are conducted in this paper.The studies of this paper include no-reference quality evaluation of underwater images,underwater image sharpening,and underwater image stitching.The research achievements are as follows:(1)In order to solve the problem of lacking in high-accuracy and robust underwater image quality evaluation methods,a no-reference quality evaluation metric specially used for underwater images is proposed in this paper.Firstly,based on the perception characteristics of the human visual system and the transmission characteristics of light in water,the colorfulness index,contrast index,and sharpness index are designed to perceive the changes of underwater image quality.For the contrast index and sharpness index,an adaptive grayscale conversion method is proposed to enhance their generalization ability of color.Finally,the proposed evaluation metric is obtained by a linear combination of the above three indexes,and the evaluation results of the proposed evaluation metric are more accurate than other underwater image quality evaluation metrics.(2)In order to solve the problem of low contrast and color distortion in underwater optical images,an underwater image restoration algorithm base on the underwater imaging model and convolutional neural network is proposed in this paper.Firstly,the underwater imaging model is transformed.Two unknown parameters,background light and transmission of the underwater imaging model,are integrated into one unknown parameter.Then a new multi-scale convolution neural network is designed to estimate the only one unknown parameter of the transformed underwater imaging model.Finally,the restored image is obtained by the inversion of the underwater imaging model.The results of comparison experiments show that the proposed restoration algorithm achieves a better processing effect than other underwater image sharpening algorithms.(3)In order to solve the problem that the view of a single image is limited by the short visual distance of underwater optical cameras,a fast stitching algorithm used for multiple underwater images is proposed in this paper.The proposed stitching algorithm completes the continuous stitching of multiple underwater images by preprocessing,feature point extraction and matching,overlapping area alignment,stitched image optimization,and stitching strategies of multiple images.In the stage of feature point matching,a feature point matching algorithm based on multiple matching strategies is proposed.It achieves high-precision feature matching by the combination of Euclidean distance threshold,vertical spatial distance threshold,and approximate overlap width condition.In the process of stitching multiple images,a stitching strategy based on the region of interest extraction is proposed.It effectively improves the efficiency of stitching by disassembling images.