光学学报, 2017, 37 (4): 0410001, 网络出版: 2017-04-10
针对动态目标的高动态范围图像融合算法研究 下载: 641次
High Dynamic Range Image Fusion Algorithm for Moving Targets
图像处理 动态目标 色彩梯度微分光流法 高动态范围图像融合权重函数 image processing moving target color-gradient-based differential optical flow met high dynamic range image fusion weighting function
摘要
高动态范围成像技术能够全面有效地反映拍摄场景信息,有利于在复杂环境下获得更高的成像质量。目前基于多曝光图像序列的高动态范围图像融合算法需要高精度配准的输入图像序列,无法克服动态问题带来的影响,无法针对存在动态目标的多曝光图像序列进行高动态范围图像融合。基于此,提出了一种高动态范围图像融合算法。该算法利用基于色彩梯度的微分光流法获得由相机抖动以及场景目标运动导致的多曝光图像之间的动态目标偏移量;结合标定的逆相机响应函数构建高动态范围图像融合权重函数,对存在动态问题影响的多曝光图像进行高动态范围图像融合。实验结果表明,提出的算法无需提前对输入图像序列进行精确配准,能够有效克服动态问题影响,实现动态目标的高动态范围图像融合。
Abstract
High dynamic range imaging technique can reflect the shooting scenario comprehensively and effectively, resulting in high quality imaging in the complex environment. However, the classic high dynamic range image fusion algorithm based on high-precision registration of multiple exposure images cannot deal with the impact of dynamic problems. It fails to realize high dynamic range image fusion when there is a moving target in the exposure images. Therefore, a new high dynamic range fusion algorithm to deal with the dynamic targets is proposed. The derivative optical flow based on color gradient is first utilized to acquire the dynamic target offset of different exposure images, which is caused by the camera shark or the movement of the scenario target. The high dynamic range image fusion weighting function is established by combination of target offset and inverse camera response function, and is applied to the fusion algorithm to tackling the problem of dynamic targets in the high dynamic range images without precise mapping. The experimental results show that without image registration, the proposed algorithm is effective in the fusion of the high dynamic range images with moving targets.
都琳, 孙华燕, 王帅, 高宇轩, 齐莹莹. 针对动态目标的高动态范围图像融合算法研究[J]. 光学学报, 2017, 37(4): 0410001. Du Lin, Sun Huayan, Wang Shuai, Gao Yuxuan, Qi Yingying. High Dynamic Range Image Fusion Algorithm for Moving Targets[J]. Acta Optica Sinica, 2017, 37(4): 0410001.