Research On Object-oriented SAR Image Enhancement Algorithms

Synthetic Aperture Radar(SAR),with its all-day,all-weather and penetrating capabilities,has been widely used in military,production and livelihood fields and has shown strong application potential.The raw SAR images need to be analyzed by professionals or professional software in order to obtain meaningful results,so the quality enhancement of SAR images is both practical and important.In this paper,we focus on the SAR image enhancement algorithm and its implementation,as follows:Aiming at the shortcomings of the traditional partial differential equations algorithms that cannot simultaneously maintain edge features and effectively smooth the image,a hybrid model applying both P-M nonlinear equation and coherent enhanced partial differential equation is proposed for SAR image enhancement.The detailed principles of the P-M nonlinear equation and the coherent enhancement equation are fully explained and the algorithm is programmed for implementation.The hybrid model is applied to enhance SAR images.And the quality of the enhanced image is evaluated in terms of the whole and also the details.Objective evaluation metrics are introduced to compare the images before and after enhancement,and the effectiveness of this hybrid partial differential equation model was verified by visually discriminating the overall image brightness as well as the specific target detail enhancement,a 25% increase in image mean value,and a small increase in information entropy.In order to solve the problem that between the reconstruction effect and the convergence speed in the super-resolution reconstruction process,one can often only discard one and pursue the other,an improved adaptive line search strategy is used to iterate the model.The process of generating the bilateral total variance(BTV)regularization method and its detailed principles are explained.The improved iterative method is used to achieve fast convergence of this super-resolution reconstruction algorithm,accelerate the SR process,and obtain the optimal solution with fewer iterations.The selected objective evaluation metrics performed well,in which the point sharpness function EAV2 improved by a factor of 20 times compared to the original images,the equivalent number of looks(ENL)improved by nearly 50%.On the other hand,the program running time was reduced by half compared to the traditional fastest gradient descent method,illustrating the superiority of the proposed adaptive line search method.Starting from the correlation between the nonlinear process of estimating highresolution SAR images from low-resolution SAR images and the nonlinear state estimation of the unscented Kalman filter,the initial state is estimated more efficiently and accurately by the unscented transformation before super-resolution reconstruction.And the reconstruction results of the unscented Kalman method with improved initial value estimation are compared with those of the bicubic interpolation method,and the objective evaluation indexes are improved more.Among them,the structural similarity SSIM is improved by nearly 1/5,the point sharpness function EAV2 is improved by nearly2 times,and the equivalent number of looks ENL is improved by nearly 10 times due to the image becoming twice the original size,and the superiority of the improved method is verified in combination with the visual evaluation.