摘要
同一场景不同曝光的图像序列,常出现曝光不足或曝光过度的区域,造成高亮或阴暗处的细节损失。针对这一问题,提出了基于Sigmoid函数拟合的多曝光图像直接融合算法。该算法从同一场景不同曝光的LDR图像序列中提取每个像素位置亮度信息,利用视觉适应的S形曲线,建立亮度序列曲线的数学模型,给出像素最佳成像亮度值判别方法,避免了HDR图像重建和色调映射的复杂计算,快速合成直接在常规设备上显示的HDR图像。实验结果表明,该方法具有更高的计算效率,极大程度地保留了图像细节,并对运动目标和相机微小移动的影响具有良好的鲁棒性。
In the image sequences of the same scene with different exposure values,underexposed or overexposed areas often appear,which causes the loss of the details in highlight or shadow areas. In order to solve this problem,a direct fusion algorithm is proposed for multi-exposed images based on sigmoid function fitting. First of all,the intensity information of each pixel is extracted from the low dynamic range（ LDR） image sequence of the same scene with different exposure values,and then using the s-shaped curve adapted to human vision,a mathematic model of the brightness sequence curve is built. The method for determining the pixel best imaging luminance value is given,which can avoid the complex computation for high dynamic range image reconstruction and tone mapping and synthesize the HDR image directly shown on the conventional display device rapidly. Sufficient experiments on various scenes indicate that the proposed method has higher calculation efficiency,greatly retains the image details and has good robustness to the influence of moving object and tiny displacement of the camera.
出处
《仪器仪表学报》
EI
CAS
CSCD
北大核心
2015年第10期2321-2329,共9页
Chinese Journal of Scientific Instrument
基金
陕西省教育厅自然科学基金(14JK1014)项目资助
关键词
高动态范围图像
图像融合
多曝光
曲线拟合
high dynamic range image
image fusion
multi-exposures
curve fitting
