摘要
针对聚焦形貌恢复技术在测量运行时对上位机内存需求量大的缺陷,提出一种聚焦形貌恢复算法。该算法基于聚焦评价曲线的区间特性及连续表面约束,运用传统方法初始化扩散区域,利用已扩散区域的深度信息猜测扩散点的深度,并在搜索半径内重新搜索确认扩散点的深度,得到更加精确的深度值。实验结果表明,与传统方法相比,该算法在保证精度的同时,上位机内存使用率平均降低9.6%。
Shape from Focus（ SFF） is a layer scanning technology,which reguires a lot of memory of upper computer when mehshring. To solve this problem,an algorithm based on spreading is proposed. Based on the interval property of the focus evaluation curve and the constraint of continuous surface,the algorithm initializes the spreading region using traditional methods,infers the depth of the spreading point utilizing the depth information of the spreading region,searches and confirms the depth of the spreading point within the search radius,and gets the more accurate depth value.Experimental result shows the proposed algorithm can reduce the inemory utilization by 9. 6% averagely while ensuing the precision of shape recovery.
出处
《计算机工程》
CAS
CSCD
北大核心
2016年第3期259-265,共7页
Computer Engineering
基金
西南科技大学研究生创新基金资助项目(13ycjj42)
关键词
聚焦形貌恢复
聚焦评价曲线
扩散
区间特性
上位机内存
Shape from Focus（SFF）
focus evaluation curve
spreading
interval property
memory of upper computer
