CAJ | 학술논문

针对激光主动偏振图像的散斑去除问题，提出了一种基于异向扩散的冲击各向异性滤波模型。该模型融合了冲击滤波器和小核值相似区算法，利用小核值相似区算法提取偏振图像边缘，减少了噪声对边缘检测的影响；针对不同的图像区域，自动调整冲击滤波器的系数，使得该算法既能保持图像边缘，又可以很好地抑制图像的散斑。使用八方向一阶差分估计小核值相似区算法的门限，门限估计更加准确；迭代终止条件采用完全散射区域的平均绝对误差作为标准，使得迭代次数更加合理。通过对比等效视数和边缘保持指数可见，冲击各向异性滤波算法的散斑抑制能力和边缘保持能力与传统的Lee和SRAD模型相比更加有效。
침대격광주동편진도상적산반거제문제，제출료일충기우이향확산적충격각향이성려파모형。해모형융합료충격려파기화소핵치상사구산법，이용소핵치상사구산법제취편진도상변연，감소료조성대변연검측적영향；침대불동적도상구역，자동조정충격려파기적계수，사득해산법기능보지도상변연，우가이흔호지억제도상적산반。사용팔방향일계차분고계소핵치상사구산법적문한，문한고계경가준학；질대종지조건채용완전산사구역적평균절대오차작위표준，사득질대차수경가합리。통과대비등효시수화변연보지지수가견，충격각향이성려파산법적산반억제능력화변연보지능력여전통적Lee화SRAD모형상비경가유효。
In order to reduce speckle noise of laser active polarization image, a shock anisotropic denoising model was proposed. The model utilized Smallest Univalue Segment Assimilating Nucleus (SUSAN) algorithm to extract image edge which reduced the noise influence on edge detection influence. The approach adjusted coefficient of the shock filter automatically, which made new algorithm both retain the image edge, and restrain image speckle. The threshold of SUSAN was estimated by eight direction order difference, which maked the estimation more accuracy. Fully developed regional average absolute error was used as the standard of iterative termination conditions. Through comparing the equivalent numbers of looks (ENL) and edge preserve index (EPI), the proposed algorithm provides more effective speckle reduction as well as edge preservation.