Cellular Neural Networks Based On Variation Of Partial Differential Equations Committed To Underwater Image Noise Reduction

The twenty-first century will be the century of the ocean, vast sea area is rich in resources. This will alleviate the growing population, resources, and environmental pressure. Because of the complexity of the marine environment, the process of developing marine faces a number of difficulties. In the marine environment, the sea water has very strong absorption of light attenuation, scattering, while there are a lot of marine lives, especially in small plankton and other small particles, which results in low signal to noise ratio of underwater images, poor contrast, and details fuzzy, an effect of atomization of the whole image. Therefore, effective removal of underwater image noise, clearly reflecting the information, plays an important role on the development of ocean.As the water and suspended particles absorb and scatter light, making underwater image blurring and observed degraded, which is captured by imaging system. The illumination light of the system generated backscattering in the imaging space, and the backscattering forms noise background in the image formation, which is the main reason that cause reduced image SNR and low contrast. Cellular neural networks have the advantages of local connections, linear and segmented output function, the characteristics of real-time signal processing, and can achieve the goal of large scale integrated circuits, high-speed parallel processing and so on. Image processing based on partial differential equations method has a strict theoretical basis, so it can realize image nonlinear filtering, image noise suppression, while protecting the edge information. This paper combines the advantages of neural network and PDE for image processing, to study the application of this method in underwater images backscattering noise reduction.Firstly, this paper analyzes the optical theory of underwater images, and do research on underwater images of the forward scattering and backscattering characteristics through experiments, focus on theoretical analysis of the underwater image transmission function model and the backward scattering properties. Based on this model, we give a new idea of image restoration based on the physical mechanism.Secondly, do study on image processing, which is based on partial differential equations, analysis several major variations of denoising model and the advantages and disadvantages through experimental and theoretical study, then introduce the concept of cellular neural network model and illustrate various parameters of cellular neural networks, analysis the dynamic and stability of cellular neural network, focus on cellular neural networks based on image processing principles. We give a few common gray image processing methods and results through experimental instructions.Finally, I do research on cellular neural networks based on variation of partial differential equations committed to underwater image noise reduction method. First, use cellular neural networks with typical partial differential equations to do underwater image processing. Experiments show that the isotropic diffusion:heat diffusion, Laplace equation can suppress the underwater scattering noise, but the effect is not obvious, the improved heat diffusion equation has better effect. using cellular neural networks based on the Poisson equation to do underwater image processing, having better results. So underwater image degradation process and the Poisson equation has similarities. P-M method of anisotropic diffusion denoising results is poor. Because edge detection operator seriously affected by backscattering noise. Then cellular neural networks based the general regularization method can be very good at underwater image backscattering noise reduction. Focus on cellular neural networks using variational methods, in underwater scattering images and underwater images processing. Experiment analysis of the four kinds of variation model in underwater images backscattering noise reduction. Use the parameters and power spectral density analysis of denoising results. Show that algorithm 3 can better suppress underwater images backscatter noise, and well protected image edge information. Underwater image transfer function model in this paper is reasonable. Type (5-29) describes the energy function more in line with the real underwater image energy function. The twenty-first century will be the century of the ocean, vast sea area is rich in resources. This will alleviate the growing population, resources, and environmental pressure. Because of the complexity of the marine environment, the process of developing marine faces a number of difficulties. In the marine environment, the sea water has very strong absorption of light attenuation, scattering, while there are a lot of marine lives, especially in small plankton and other small particles, which results in low signal to noise ratio of underwater images, poor contrast, and details fuzzy, an effect of atomization of the whole image. Therefore, effective removal of underwater image noise, clearly reflecting the information, plays an important role on the development of ocean.As the water and suspended particles absorb and scatter light, making underwater image blurring and observed degraded, which is captured by imaging system. The illumination light of the system generated backscattering in the imaging space, and the backscattering forms noise background in the image formation, which is the main reason that cause reduced image SNR and low contrast. Cellular neural networks have the advantages of local connections, linear and segmented output function, the characteristics of real-time signal processing, and can achieve the goal of large scale integrated circuits, high-speed parallel processing and so on. Image processing based on partial differential equations method has a strict theoretical basis, so it can realize image nonlinear filtering, image noise suppression, while protecting the edge information. This paper combines the advantages of neural network and PDE for image processing, to study the application of this method in underwater images backscattering noise reduction.Firstly, this paper analyzes the optical theory of underwater images, and do research on underwater images of the forward scattering and image transmission function model and the backward scattering properties. Based on this model, we give a new idea of image restoration based on the physical mechanism.Secondly, do study on image processing, which is based on partial differential equations, analysis several major variations of denoising model and the advantages and disadvantages through experimental and theoretical study, then introduce the concept of cellular neural network model and illustrate various parameters of cellular neural networks, analysis the dynamic and stability of cellular neural network, focus on cellular neural networks based on image processing principles. We give a few common gray image processing methods and results through experimental instructions.Finally, I do research on cellular neural networks based on variation of partial differential equations committed to underwater image noise reduction method. First, use cellular neural networks with typical partial differential equations to do underwater image processing. Experiments show that the isotropic diffusion: heat diffusion, Laplace equation can suppress the underwater scattering noise, but the effect is not obvious, the improved heat diffusion equation has better effect, using cellular neural networks based on the Poisson equation to do underwater image processing, having better results. So underwater image degradation process and the Poisson equation has similarities. P-M method of anisotropic diffusion denoising results is poor. Because edge detection operator seriously affected by backscattering noise. Then cellular neural networks based the general regularization method can be very good at underwater image backscattering noise reduction. Focus on cellular neural networks using variational methods, in underwater scattering images and underwater images processing. Experiment analysis of the four kinds of variation model in underwater images backscattering noise reduction. Use the parameters and power spectral density analysis of denoising results. Show that algorithm 3 can better suppress underwater images backscatter noise, and well protected image edge information. Underwater image transfer function model in this paper is reasonable. Type (5-29) describes the energy function more in line with the real underwater image...