Research On SAR Image Filtering Algorithm And New Evaluation Model

Synthetic aperture radar(SAR)is an active remote sensing observation system.The electromagnetic waves the SAR irradiates have unique character,so that the imaging mechanism is not affected by conditions such as light and weather.And it can achieve earth observation all-weather all day long,widely used in military and civilian field.However,the coherent imaging mechanism of radar system is the main cause of coherent speckle noise in SAR images.The speckle noise seriously damages the information of SAR image,which makes the application of SAR image very difficult.Therefore,it is necessary to suppress the speckle with the filtering algorithm and evaluate the performance of the filtering algorithm is necessary in the process of SAR image processing.In this paper,a novel filtering algorithm for the real SAR image is proposedbased on normalized soft morphological and parameterized symmetric logarithmic image processing model.Speckle noise makes the SAR image to be highlighted.Therefore,we define the distinction criterion of high and low brightness region.According to the characteristic of parameterized symmetric logarithmic image processing model which establishes a parameterized symmetric structure processing parts of the image and the flexibility of the order-statistic soft morphological operations,the high volatility in SAR image is processed.The details and edge information are also enhanced.Finally,the algorithm can not only reduce the speckle noise by morphological operation for the low brightness area but also enhance the structure and edge information of the image.This paper comparedthe proposed filtering algorithmwith some classical filtering algorithms,and the time complexity of the proposed algorithm is analyzed.Experimental results show that the proposed algorithmstrongly suppress speckle while preserving image details andsharpness and consumes less time,which is consistent with the judgment result of human visual system.Secondly,this paper establishes a new structure correlation non-reference model for filtering algorithm,analyzes propertiesof the model and prove them.By using the weight coefficients,ratio images,and structural correlations between SAR images and ratio images,this modelovercomes the problem that traditional evaluation indicators require ideal noiseless images as reference images while evaluating SAR image filtering algorithms.Therefore,the value of w in the expression weighs,through a convex sum,which makes it possible to set different weight values from a practical point of view,and verify the consistency of the evaluation model.When the weight coefficient is in [0,0.5(8),the evaluation model focuses more on evaluating the texture,structure,and detail information protection capabilities of SAR image filtering algorithms.Besides,when the weight coefficient is in(7)0.5,1],the evaluation model focuses more on evaluating the speckle noise suppression capabilities of SAR image filtering algorithms.Obviously,when comparing different despecking filters,the w must be the same for the sake of a fine comparision.Experimental results show that the new evaluation model proposed in this paper can evaluate the performance of SAR image filtering algorithms more objectively and comprehensively,and achieve the consistency of subjective and objective evaluation.