| 摘要: |
| 梯度域图像融合通过最小化融合图像的梯度场与融合梯度场之间距离求得融合图像，融合梯度场的准确性直接决定融合图像质量；由于源图像信息之间存在互补性，基于结构张量的梯度融合方法得到的融合梯度场的准确性不高，为保证融合图像质量，重构时需要添加其它感知项进行校正；提出了一种基于边缘图的梯度场融合方法，在边缘像素处以梯度值为权重进行加权融合，在非边缘像素处采用梯度值取大规则进行融合;实验结果表明：梯度场融合方法得到的融合梯度场准确性高，重构的融合图像具有满意的视觉质量和客观评价成绩。 |
| 关键词: 图像融合 梯度场 结构张量 边缘图 感知项 |
| DOI： |
| 分类号: |
| 基金项目: |
|
| Research on Edge Map Based Image Fusion Method in Gradient Domain |
|
YANG Hui jun
|
| Abstract: |
| Fusion image is solved by gradient image fusion via minimizing the distance between the gradient field and fusion gradient field of fusion image，therefore，the accuracy of fusion gradient field directly determines the quality of fusion image.Because there is the complementarity between the information of source images，the accuracy of the structure tensor based gradient field is not high.In order to ensure the quality of fusion image，other perception terms are needed to be added to correct while reconstruction.Thus，this paper advocates an edge map based gradient field fusion method，weighting fusion is processed at edge pixel by the weight of gradient value while maximum modules are fused at non edge pixel.Experiment result indicates that the accuracy of the gradient field by the gradient field fusion is high and that the reconstructed fusion images have satisfactory visual quality and objective assessment quality. |
| Key words: image fusion gradient field structure tensor edge image perception term |
