【摘要】 同一场景的多曝光图像序列被广泛的应用于高动态范围图像(HighDynamicRangeImage)的合成中。但是,在多曝光图像序列的采集过程中,相机抖动、场景运动等因素会对合成图像的质量产生较大的影响。此外,离镜头较近的大目标往往由于显著的三维形状,在序列图中产生较大的视差效应,也会对合成图像产生消极影响。该文提出一种基于SURF特征点的三维图像配准算法,实验证明该算法在近距离大目标情形下较之传统配准算法MTB(MeanThresholdBitmap,均值二值化)可以获得更好效果。更多还原

【Abstract】 Multi-exposure images have been widely used for High Dynamic Range Image composition. However,camera vibration or scenemovement can undermine the fusing result severely. Besides,objects with significant 3D structures can often introduce parallax error in registration.Therefore,an effective and robust registration technique is crucial before the HDR composition process. An improved HDR image registration method using SURF is proposed in this paper.The experiment shows the effectiveness of this method. Making a comparison of method to the MTB(mean threshold bitmap),in scenes which contains significant 3D structures,the algorithm shows better registration results and have much smaller blurring phenomenon after HDR fusing.更多还原