基于金字塔变换的多曝光图像融合

Multi-exposure Image Fusion Based on Pyramid Transformation

在同一场景不同光线下得到的图像序列,常出现过曝光或曝光不足的现象,造成图像质量低,重要信息丢失。为解决这些问题,文中对拉普拉斯金字塔与梯度金字塔变换融合的方法进行了分析研究,后者增加了图像方向信息。在图像融合过程中,采用分块的方法,首先生成原始多曝光图像的权重图,再根据金字塔原理对权重图进行加权平均得到融合后图像。实验结果表明,该算法能够得到具有更多有用信息的高对比度的融合图像,融合效果良好,具有较好的实用价值。

The images which obtained in the same scene under different light are often over exposed and underexposed. These image have low quality. Lots of important information is loss. To solve these problems, the Laplacian pyramid and gradient pyramid transformation are analyzed. The latter increases the information of image direction. In the process of image fusion, using the block method, it is first to gen- erate the weight diagram of the original multi-exposure images. Then generate a high dynamic range images based on Laplacian pyramid weighted average of the weight figure. The experimental results show that the algorithm can obtain high-contrast fusion image with more useful information. The algorithm works well and has good practical value to some extent.