基于改进特征光流估计的多曝光图像融合

MULTI EXPOSURE IMAGE FUSION BASED ON FEATURE OPTICAL FLOW ESTIMATION

  • 摘要: 针对多曝光图像拍摄过程中目标对象局部运动导致融合图像存在鬼影现象的问题,提出一种对亮度变化具有鲁棒性的色彩方向梯度直方图特征,并结合图像灰度信息、结构信息和颜色信息构建一种基于改进特征光流估计的多曝光图像融合算法。实验结果表明:与现有算法相比,高动态范围图像评价指标pu_psnr提升了7.5dB,log_psnr提升了7.5dB,pu_ssim提高了5%。

     

    Abstract: In order to solve the problem of ghost phenomenon in the fused image caused by the local movement of the target object in the process of multi exposure image shooting, a color direction gradient histogram feature that is robust to brightness changes was proposed, and a multi exposure image fusion algorithm based on improved feature optical flow estimation was constructed by combining the image grayscale information, structural information, and color information. The experimental results show that compared with the existing algorithms, the high dynamic range image evaluation index pu_psnr has increased by 7.5 dB, log_psnr is increased by 7.5 dB, pu_ssim index is increased by 5%.

     

/

返回文章
返回