摘要
光场相机获取的四维光场信息可用于场景深度估计,但是大多深度估计算法获得的深度图存在边缘模糊、精度有限等问题。因此,结合引导滤波器边缘保持局部平滑特性,提出基于纹理信息引导的光场深度图优化算法。该算法以纹理信息丰富的光场中心孔径图像作为引导图像,建立了基于多评价函数的混合引导滤波参数寻优模型,以获得合理滤波器参数实现深度图引导滤波优化。实验结果表明,优化后的深度图边缘的视觉效果明显改善,与散焦结合相关性评价算法获得的原始深度图相比,均方误差平均降低1.12%。
The four dimensional information captured by light field camera can be used for depth estimation.The edge of the depth maps calculated by most depth estimation algorithms have the problem of artifact and limited accuracy.Based on the feature of edgepreserving smoothing of guided filtering,a light field depth map optimization algorithm is proposed based on texture image guided filtering.Using the center sub-aperture image of the light field as the guided image,a mixture filtering parameter optimization model is established and the accuracy of the depth map is improved.Experimental results demonstrate the effectiveness of the proposed method.The visual effect of the optimized depth map is improved obviously.Compared with the initial depth maps from combining defocus and correspondence,the mean square error of the optimized depth map is reduced by 1.12%on average.
作者
武迎春
赵志浩
王玉梅
王安红
赵贤凌
WU Yingchun;ZHAO Zhihao;WANG Yumei;WANG Anhong;ZHAO Xianling(School of Electronic and Information Engineering,Taiyuan University of Science and Technology,Taiyuan 030024,China)
出处
《电视技术》
2020年第7期51-55,共5页
Video Engineering
基金
国家自然科学基金青年基金(61601318)
山西省回国留学人员科研资助项目(2020-128)
山西省应用基础研究项目(201601D021078)
山西省重点学科建设经费
山西省互联网+3D打印协同创新中心
山西省1331工程重点创新团队
山西省科技创新团队(201705D131025)
太原科技大学博士启动基金(20132023,20192072)
国家留学基金。
关键词
深度图优化
引导滤波
纹理信息
参数优化
优化模型
depth maps optimization
guided filtering
texture information
parameter optimization
optimization model
