Publication details

<< back
 

Generation of Synthetic Image Datasets for Time-Lapse Fluorescence Microscopy

Basic information
Original title:Generation of Synthetic Image Datasets for Time-Lapse Fluorescence Microscopy
Authors:David Svoboda, Vladimír Ulman
Further information
Citation:SVOBODA, David and Vladimír ULMAN. Generation of Synthetic Image Datasets for Time-Lapse Fluorescence Microscopy. In Campilho, Aurélio; Kamel, Mohamed. Proceedings of 9th International Conference on Image Analysis and Recognition. LNCS 7325, Part II. Heidelberg: Springer-Verlag, 2012. p. 473-482, 10 pp. ISBN 978-3-642-31297-7. doi:10.1007/978-3-642-31298-4_56.Export BibTeX
@inproceedings{974481,
author = {Svoboda, David and Ulman, Vladimír},
address = {Heidelberg},
booktitle = {Proceedings of 9th International Conference on Image Analysis and Recognition},
doi = {http://dx.doi.org/10.1007/978-3-642-31298-4_56},
edition = {LNCS 7325, Part II},
editor = {Campilho, Aurélio; Kamel, Mohamed},
keywords = {Simulation; Optical flow; 3D image sequences; Fluorescence optical microscopy},
howpublished = {tištěná verze "print"},
language = {eng},
location = {Heidelberg},
isbn = {978-3-642-31297-7},
pages = {473-482},
publisher = {Springer-Verlag},
title = {Generation of Synthetic Image Datasets for Time-Lapse Fluorescence Microscopy},
year = {2012}
}
Original language:English
Field:Use of computers, robotics and its application
Type:Article in Proceedings
Keywords:Simulation; Optical flow; 3D image sequences; Fluorescence optical microscopy

In the field of biomedical image analysis, motion tracking and segmentation algorithms are important tools for time-resolved analysis of cell characteristics, events, and tracking. There are many algorithms in everyday use. Nevertheless, most of them is not properly validated as the ground truth (GT), which is a very important tool for the verification of image processing algorithms, is not naturally available. Many algorithms in this field of study are, therefore, validated only manually by an human expert. This is usually difficult, cumbersome and time consuming task, especially when single 3D image or even 3D image sequence is considered. In this paper, we have proposed a technique that generates time-lapse sequences of fully 3D synthetic image datasets. It includes generating shape, structure, and also motion of selected biological objects. The corresponding GT data is generated as well. The technique is focused on the generation of synthetic objects at various scales. Such datasets can be then processed by selected segmentation or motion tracking algorithms. The results can be compared with the GT and the quality of the applied algorithm can be measured.

Related projects:

 
 
top