Underwater Image Processing Of Low Level Vision

Underwater images are easily affected by low light condition,which result the obtained images low quality and poor visibility.Also unbalanced brightness,chromatic aberration and severe noise make the following processing steps more difficult.The degraded image targeted restoration approach can dramatically highlight the image details and enhance the global contrast,which is of benefit to improve the visual effect during practical applications.Also Due to the complexity of underwater channel and various interferences,underwater sonar image resolution is low which gives rise to many difficulties to identify the edge detail.Low-level vision problems mainly contain image restoration and enhancement.Image restoration problem remains a significant field of image processing while image enhancement is an important step among pre-processing techniques and of great meaning to the following steps.The restoration approach to degraded underwater images can dramatically stress the image details and enhance image contrast,which is of benefit to improve image visual effect in practical application.This paper focuses on image processing theory and its application,which mainly includes multi-scale geometric analysis,Retinex,Markov random field(MRF),dictionary learning,sparse coding,spare regularization and their typical applications in image processing.The specific research contents are as follows:(1)We analyze various image restoration methods for different image degradation models and discuss how to apply deconvolution to match suitable blur model.On these bases,we propose new methods to realize image restoration via estimation.The proposed method considers the Radon transform and Fourier transform to estimation the unknown parameters in the Point Spread Function,which is more simple and precise comparing to other estimation methods.Simulation results show that the proposed method can greatly improve the image quality and definition.(2)The restore method aiming at Gaussian noise contaminated incoherence underwater images proposed is based on Surfacelet and Retinex.By transforming image from RGB color space into HS V color space,hue(H)component is kept the same,saturation(S)component is denoised using Surfacelet and value(V)component is enhanced by Retinex,finally the composed HSV image will be transformed back to RGB color space and the restored image can be acquired.Simulation results show that the proposed algorithm has its own advantages on edge detail maintenance and color reservation and is beneficial to the subsequent image processing research.(3)In order to restore the impulse noise contaminated incoherence underwater images,the pulse coupled neural networks(PCNN)based homomorphic enhancement method has been proposed.The proposed method meanwhile takes human visual characteristic into consideration and the output results turn out to be improved and the detail vivid.The wavelet transform has been introduced into original homomorphic filtering and then the PCNN enhancement has been conducted.Compared with other methods,the simulation applied demonstrates the proposed method effective.(4)Fields of Experts(FoE)based coherence image restoration has been discussed and some open issues including noise estimation,parameter selection have been approached.The stochastic method FoE performs fairly well,meanwhile it might also produce unsatisfactory outcome especially when the noise is grave.To improve the performance we introduce the deterministic method K-SVD.The FoE-treated image has been used to obtain the dictionary,and with the help of sparse and redundant representation over trained dictionary,the K-SVD algorithm can dramatically solve the problem even the pre-treated result is of poor quality under severe noise condition.The experimental results via our proposed method are demonstrated and tested in detail.Meanwhile comparative results from both qualitative and quantitative are given to present the better performance over current state-of-art related restoration algorithms.(5)Based on color-consistent Retinex theory and aiming at improve the images acquired in underwater environment,an image enhancement method consisted of histogram and Retinex has been proposed.The decomposed image components will be enhanced by both methods and the result of higher entropy will be reserved and composed into the enhanced image.Contrastive simulation experiments have been conducted and compared with single histogram equalization or Retinex method,the proposed method owns better performance and visual effect.(6)Based on Dark Channel Prior principle,a serious of improvements have been made to estimate the transmission map of fuzzy underwater images.The total variation and image morphology methods(specifically top-hat transform and bottom-hat transform)have been introduced to the improved method.Compared with original transmission map estimation method,the proposed owns both simplicity and accuracy.The estimated transmission map together with the element can enhance the fuzzy image.Simulation results show that the processing can greatly improve the image quality and definition.