| 摘要: |
| 提出了一种直接从同一场景多次不同曝光值下成像的LDR(low dynamic range)图像序列中提取每个像素位置最佳成像信息的图像融合方法,可以在无需任何拍摄相机参数及场景先验信息的情况下,快速合成适合在常规设备上显示的HDR(high dynamic range)图像.该方法利用特殊设计的鲁棒性曲线拟合算法建立LDR 图像序列中每个像素位置像素值曲线的数学模型,并由此给出评价单个像素成像时曝光合适程度的标准和融合最佳成像像素信息的方法.对不同场景的大量实验结果显示,该方法的计算结果与传统HDR 成像技术经过复杂的HDR 重建和色调映射计算后得到的结果相当,但具有更高的计算效率,并同时对图像噪声、相机微小移动和运动目标的影响具有较好的鲁棒性. |
| 关键词: 高动态范围成像 鲁棒性曲线拟合 图像融合 最小二乘拟合 |
| DOI：10.3724/SP.J.1001.2011.03795 |
| 分类号: |
| 基金项目:国家高技术研究发展计划(863)(2007AA01Z328, 2009AA01Z332) |
|
| Directly Fusion Method for Combining Variable Exposure Value Images |
|
ZHANG Jun1,2, DAI Xia2, SUN De-Quan1, WANG Bang-Ping1
|
|
1.Institute of Image and Graphic, College of Computer Science, Sichuan University, Chengdu 610064, China;2.Institute of Computer and Mathematics, Xihua University, Chengdu 610039, China
|
| Abstract: |
| In this paper, an efficient and robust image fusion method is proposed to extract pixel intensity under the best exposure value directly from a low dynamic range of images with variable exposure values. Using this method, the HDR (high dynamic range) image can be computed and shown on the traditional display device without any prior information on a camera or scene. The proposed image fusion method is based on a mathematical model of a pixel intensity curve, established by a special robust curve fitting arithmetic criteria to evaluate the image quality of each pixel. The sufficient experimentations on various scenes indicate that the proposed method is more efficient than traditional HDRI (high dynamic range imaging) technologies and more robust when imaging noise, object movement, and small camera displacement simultaneously. |
| Key words: high dynamic range imaging robust curve fitting image fusion least squares fitting |
