CAJ | 학술논문

由于毫米波图像分辨率低并伴随有大量噪声，其图像边缘常被噪声污染或丢失，为此提出保边缘自蛇模型并应用于毫米波图像去噪。引入只在图像的边缘处具有最大值，且对噪声不敏感的非局部梯度，以此构造边缘停止函数，使其在图像边缘处接近0，而在平坦同质区域接近1；给出保边缘的自蛇模型迎风差分数字解法；提出毫米波图像去噪的定量评价算法性能指标。实验结果表明，该算法在等效视数和边缘清晰度性能指标上明显优于标准自蛇模型和保特征的各向异性扩散模型，具有较好的去噪效果和保边缘能力。
유우호미파도상분변솔저병반수유대량조성，기도상변연상피조성오염혹주실，위차제출보변연자사모형병응용우호미파도상거조。인입지재도상적변연처구유최대치，차대조성불민감적비국부제도，이차구조변연정지함수，사기재도상변연처접근0，이재평탄동질구역접근1；급출보변연적자사모형영풍차분수자해법；제출호미파도상거조적정량평개산법성능지표。실험결과표명，해산법재등효시수화변연청석도성능지표상명현우우표준자사모형화보특정적각향이성확산모형，구유교호적거조효과화보변연능력。
For millimeter wave image having a lower resolution and a lot of noises, edges are partly lost to a certain extent or contaminated severely by noise. In this paper, a new edge-preserving self-snake model for the millimeter wave image de-noise is presented. The edge stopping function which is approximated to zero around fine structures and is close to one in the interior of homogeneous regions is constructed with the non-local gradient having maximum peak only at edges and good tolerance for noise. A discrete algorithm for proposed model is given with upwind differential scheme. The quantitative evaluation of performance is also defined. In visual quality and the performance of ENL(Equivalent Number of Looks) and edge definition, the experiment results show that this algorithm is better than conventional self-snake model, and feature-preserving nonlinear anisotropic diffusion.