基于变分法与纹理分解的SAR图像分割与目标检测研究

Based on the variational method and texture information, the project consider SAR image segmentation and target detection. The main contents include: combining the active contour level set, non-local operator and the SAR image texture, geometry topology and other features. We propose image segmentation model for the strong speckle noise and texture image and unsupervised segmentation model for interesting objects,and take advantage of the split Bregman and Chambolle projection technology to study the fast and efficient algorithms; using variational method and a suitable texture space, we propose image decomposition model based on variational PDE, and then based on texture information propose the segmentation algorithm for SAR images. Through the research, the purpose of this project is to obtain a series of SAR image segmentation algorithms and develop the PDE on the application of the SAR image segmentation. Image segmentation is an important aspects in the SAR image processing, and obviously PDE is an important and efficient technique. The project is to promote the combination of many subjects, which can solve many problems in the SAR image segmentation and propose a theoretical framework. It not only has great practical value, but also has important theoretical value and scientific significance.

本项目旨在研究基于变分法与纹理分解的SAR图像分割与目标检测，主要内容包括：结合活动轮廓水平集技术、非局部算子和SAR图像的纹理、几何拓扑等特征信息，提出强相干斑噪声的图像分割模型、强纹理图像的图像分割模型、以及感兴趣物体的无监督分割模型，利用分裂Bregman技术和Chambolle投影技术研究其高效快速算法；利用变分法和构造合适的纹理特征函数空间，建立基于变分PDE的图像分解模型和基于纹理信息的SAR图像分割算法。通过本项目的研究，我们将得到一系列SAR图像分割算法，发展变分偏微分方程在SAR图像分割与分解中的应用。图像分割在SAR图像理解的过程中是非常重要的环节，而变分偏微分方程则是图像处理的一种有效手段。因而，本项目的研究是对多学科交叉融合的促进，它即能解决SAR图像分割中的很多问题，又能够建立统一的理论框架，不仅具有很大的实用价值，而且具有重要的理论价值和科学意义。

本项目主要致力于研究基于变分法与PDE的SAR图像恢复、分割与目标检测，主要内容包括：基于变分和PDE的SAR图像相干斑去噪模型，基于非线性扩散方程的SAR图像相干斑去噪模型，变分模型和PDE模型相应的数学理论研究，SAR图像感兴趣物体检测模型，SAR图像分割快速算法。上述研究不仅可以利用数学理论解决SAR图像的恢复、分割和目标检测中的实际问题，亦可促进多学科交叉融合。