RECURSIVE FILTERS FOR OPTICAL-FLOW

被引：90

作者：

FLEET, DJ ^{[1
]}

LANGLEY, K ^{[1
]}

机构：

[1] UCL, DEPT PSYCHOL, LONDON, ENGLAND

来源：

IEEE TRANSACTIONS ON PATTERN ANALYSIS AND MACHINE INTELLIGENCE | 1995年 / 17卷 / 01期

关键词：

D O I：

10.1109/34.368151

中图分类号：

TP18 [人工智能理论];

学科分类号：

081104 ; 0812 ; 0835 ; 1405 ;

摘要：

Working toward efficient (real-time) implementations of optical flow methods, we have applied simple recursive filters to achieve temporal smoothing and differentiation of image intensity, and to compute 2d now from component velocity constraints using spatiotemporal least-squares minimization. Accuracy in simulation is similar to that obtained in the study by Barren et al. [3],while requiring much less storage, less computation, and shorter delays.

引用

页码：61 / 67

页数：7

共 24 条

[1]

Adelson E. H., 1986, Proceedings of the Workshop on Motion: Representation and Analysis (Cat. No.86CH2322-6), P151

[2]

Barman H., 1991, Proceedings of the IEEE Workshop on Visual Motion (Cat. No.91TH0390-5), P44, DOI 10.1109/WVM.1991.212789

[3] PERFORMANCE OF OPTICAL-FLOW TECHNIQUES [J].