光学技术, 2018, 44 (4): 480, 网络出版: 2018-08-30
基于图像内容评价因子的动态场景曝光融合算法
Exposure fusion algorithm in dynamic scenes based on image content assessment factor
高动态范围成像 多曝光图像融合 图像梯度 抑制鬼影 high dynamic range imaging multi-exposure image fusion image gradient ghost suppression
摘要
针对目前高动态范围(HDR)成像算法难以应对动态场景的局限性, 提出了一种有效抑制运动鬼影的HDR图像融合算法。算法根据整体图像质量, 从图像序列中选择参考图像;从像素曝光质量、细节丰富程度及与参考图像之间的局部相似性三个角度对图像内容进行评价, 根据评价结果确定融合权重;采用拉普拉斯金字塔图像融合算法进行图像融合, 得到HDR图像。采用图像熵与结构相似性(SSIM)相结合的评价标准对算法进行评价, 实验结果表明, 处理结果的平均SSIM达到0.952, 证明算法对运动鬼影具有显著的抑制效果。
Abstract
A high-dynamic-range (HDR) image fusion algorithm is proposed to deal with the poor dynamic scene performance in the current relevant algorithm. This algorithm can effectively suppress the moving ghost phenomena. According to the overall image quality, the reference image is selected from the image sequence. The image content assessment is processed to determine the fusion weights map. The assessment mainly focuses on three aspects: the pixel exposure quality, the detail richness and the local similarity with the reference image. The high dynamic range image is obtained through image fusion algorithm based on Laplacian pyramid transform. The evaluation composed of image entropy and structural similarity (SSIM) is applied to evaluate the algorithm. The experiment result indicates that the average SSIM value of output images is 0.952, which shows that the algorithm has a significant inhibitory effect on motion ghosting.
郝福得, 陈晓冬, 席佳祺, 汪毅, 齐麟. 基于图像内容评价因子的动态场景曝光融合算法[J]. 光学技术, 2018, 44(4): 480. HAO Fude, CHEN Xiaodong, XI Jiaqi, WANG Yi, QI Lin. Exposure fusion algorithm in dynamic scenes based on image content assessment factor[J]. Optical Technique, 2018, 44(4): 480.