多源影像融合与分割的协同方法

针对现有影像融合与分割方法之间缺乏协同的问题,借鉴数据同化系统能够协同其模型算子和观测算子,并且能够自适应地优化其本身的思想,提出一个多源影像融合与分割的协同框架.在该框架下,以基于对比度金字塔变换和基于非下采样的Contourlet变换的两种融合方法分别模拟模型算子和观测算子,以评价分割效果的概率随机系数为目标函数,以带交叉算子的粒子群算法作为数据同化系统的优化算法.该框架可根据融合结果影像来调整分割算法的参数,利用分割结果来指导融合结果的优化,从而使得影像融合与分割协同工作.二组实验验证了该框架的有效性.

Cooperation between Fusion and Segmentation for Multisource Image

In order to solve the problem that lack of coordination between image fusion and segmentation methods.A cooperation framework for multisource remote sensing images fusion and segmentation was proposed in view of the advantage that data assimilation system can integrate its model operator and observation operator, and it can be optimized itself.Under this framework, two fusion methods based on contrast Pyramid transform and nonsubsampled contourlet transform were used as model operator and observation operator, the objective function was composed of probabilistic rand index to evaluate segmentation effect and particle swarm optimization with crossover operator was employed.The framework can adaptively adjust the parameters of segmentation algorithm according to fused images, and can use the segmentation results to guide the optimization of fused images, so as to make image fusion and image segmentation cooperate with each other.Two groups of experiments validate the effectiveness of the framework.