| The ocean holds an important position in future survival and development.Implementing ocean strategy and developing the ocean economy cannot do without the support of intelligent underwater equipment,and optical detection technology is one of the important ways for intelligent equipment to obtain underwater environmental information.However,the propagation of light in the water is severely affected by water medium absorption,water particle scattering,and water flow fluctuations,which pose significant challenges to the feature extraction,object detection,and tracking operations of intelligent underwater devices based on visual information.Therefore,improving the quality of underwater images obtained by imaging devices through image processing technology is significant for enhancing the relevant application performance of underwater intelligent devices.Dissertation proposes a statistical-based digital contrast calculation method and three underwater image clarity methods.The main research work and innovative points are as follows:(1)A color image contrast calculation method is proposed,which first processes the original color image through two independent steps.One step is to obtain the chromaticity information of each pixel in the HSV color space;Another step is to obtain the brightness information of each pixel in the image by utilizing the linear decolorization ratio parameter value that preserves the image contrast information.Finally,the contrast value of the color image is calculated by calculating and counting the number of points whose Color difference exceeds the set threshold and the Color difference information between adjacent pixels.Experiment results show that our method is less affected by the uniform color patches than other methods.It can effectively reflect the relationship between the contrast values of different images and reduce the information loss of color contrast.(2)A method for underwater image clarity based on bright channel color compensation and scene depth estimation is proposed.This method first uses the bright channel information of the original image as the basis for color compensation to obtain a color compensated image.Then,background light information is obtained through scene depth information,and an improved background light estimation method,and the transmittance information of each pixel in the image is calculated.Finally,the estimation results are substituted into the underwater imaging model to obtain the clarity of the original underwater image.At the same time,a self-collected underwater image dataset with multiple underwater scene styles and different qualities was introduced.The experimental results show that this method can effectively complete the task of clarifying low-quality underwater images;.(3)A new method to deal with underwater images by imitating the color constancy mechanism of human vision using double-opponency.This method simulates the color constancy mechanism of human vision to achieve clarity processing of underwater images.Firstly,the original image is converted to the LMS space.Then the signals are linearly combined,and Gaussian convolutions are performed to imitate the function of receptive fields.Next,two receptive fields with different sizes work together to constitute the double-opponency response.Finally,the underwater light is estimated to correct the colors in the image.Further contrast stretching on the luminance is optional.Experiments have shown that this method can reduce the time consumption to one-sixth or less of other comparison methods while ensuring the quality of the resulting image.(4)A new underwater image clarification method involving background light estimation based on double-opponency and transmission estimation based on the red channel prior.For background light estimation,double-opponency is adopted to imitate human visual color constancy,whereby background light information consistent with human perception is obtained.Transmission estimation uses the prior knowledge that red wavelength has the fastest attenuation in water.To enhance the contrast effect of the resulting image,the restored image is fused with that of contrast stretch processing in HSV space to obtain the final output image.Experiments indicate that the proposed method can reduce computation time by 60%or more relative to comparison methods,thereby producing a clearer normal color image according to both objective and subjective evaluations. |
