Evaluation Of Quality On SAR Image Based On Textural Feature

Synthetic aperture radar (SAR) plays an extremely important role in the field of modernelectronic countermeasures. Some jamming applied in the SAR imaging process, which canhinder researchers to distinguish the SAR images. Currently, there is no uniform standard tomeasure the degree of interference suffered imaging, which to some extent limits the SARanti-jamming technology research and development. Therefore, this paper mainly studies theSAR image quality assessment method based on textural feature, makes the objectiveassessment results more accord with the subjective assessment of human visual system. Thispaper mainly completed work as follows:First of all, this paper studies the SAR imaging principle and SAR image inherenttextural feature. Using the single visual complex data of the United States Sandia laboratoryMiniSAR system, simulate the decline in quality caused several interference on SAR imagesunder conditions of specific war intelligence. The paper analyzes the advantages of thewavelet transform in the field of image processing and establishes the theoretical foundationfor the SAR image quality evaluation in the wavelet domain.Secondly, for the better use of the single complex radar images of the MiniSAR system,the method of polynomial fitting is adopted to adjust the uneven brightness which is causedby the asymmetric illumination in the process of imaging. And the paper restrains the specklenoise generated in the imaging and the technique of Lee filtering is put forward, combinedwith the adaptive weight and the adaptive window, which is approved by the experiments notonly to be suitable for the scene of uniform grayscale, but also to be suitable for the surfacefeatures scene of rather fluctuant grayscale. Then this paper focuses on several traditionalmethods of evaluation at present including the methods of the subjective evaluation, theequivalent number of looks, the mean square error, the Euclidean distance, the correlationcoefficient and the structural similarity. It proves that only the result obtained in the method ofstructural similarity accords with the objective fact, and the other methods are all equippedwith a certain limits. However, it is also should be noted that the method of structuralsimilarity is short of the description to the image textural features, so it is necessary to have animproving research to the method of structural similarity.Next, the textural features of the SAR image are utilized to deeply research the advantages and disadvantages of the three textural features of angular second moment, inversedifference moment, and contrast ratio in the image processing of SAR and the method ofsliding window block is taken to calculate the improved structural similarity of the texturalfeatures of SAR image. With the identification of the experiment, CSSIM is more accurate toevaluate quality of the disturbed SAR image. It makes use of the advantages of the wavelettransform in the domain of image processing combined the uniformity of contrast sensitivityfunction with the human visual system and proposes the method of the evaluation to the SARimage quality of the vision weight in the wavelet domain including the methods of CTWVRand CTWVR. In order to meet the needs of research in this paper, another method namedWLS is researched which also posses certain advantages in the aspect of evaluation to thequality of the SAR image. A great number of experiments demonstrate that the methodspresented by this paper meet the requirements of the image quality of the SAR image which isdisturbed; meanwhile, the methods also have certain disadvantages which prevent them beingsuitable for the disturbed images completely.Finally, the method of combining various arithmetic of the quality evaluation is also putforward in this paper. By selecting different weight coefficient based on the advantages,disadvantages and applicability of each method, the evaluation of the SAR image qualitywhich is consistent with the subjective evaluation results of human visual system is realized.