Segmentation of Natural and Man-Made Structures by Independent Component Analysis

Authors:
André Cavalcante;Fausto Lucena;Allan Kardec Barros;Yoshinori Takeuchi;Noboru Ohnishi
Affiliations:
Depto. of Media Science, School of Information Science, Nagoya University, Nagoya, Japan 464-8603;Depto. of Media Science, School of Information Science, Nagoya University, Nagoya, Japan 464-8603;Laboratory for Biological Information Processing, Universidade Federal do Maranhão, São Luís, Brazil 65080-040;Depto. of Media Science, School of Information Science, Nagoya University, Nagoya, Japan 464-8603;Depto. of Media Science, School of Information Science, Nagoya University, Nagoya, Japan 464-8603
Venue:
ICA '09 Proceedings of the 8th International Conference on Independent Component Analysis and Signal Separation
Year:
2009

Citing 5
Cited 0

What is the goal of sensory coding?

Neural Computation
Automatic object extraction from aerial imagery—a survey focusing on buildings

Computer Vision and Image Understanding
LabelMe: A Database and Web-Based Tool for Image Annotation

International Journal of Computer Vision
Man-made structure detection in natural images using a causal multiscale random field

CVPR'03 Proceedings of the 2003 IEEE computer society conference on Computer vision and pattern recognition
Delineating buildings by grouping lines with MRFs

IEEE Transactions on Image Processing

Quantified Score

Hi-index	0.00

Visualization

Abstract

Multi-scale processing is one of the main issues in the segmentation of natural and man-made structures in real worlds scenes. In this work, we use independent component analysis (ICA) to learn sets of multi-scale features specialized for natural and man-made structures, respectively. Then, we use the learned features to represent images according to a simple linear generative model. Finally, we separate each group of structures by analyzing the error of representation for each set of features. The features learned by ICA reflected both second and higher-order statistical information of each dataset. The average time consumed in the segmentation was 3 milliseconds by image block. The system was validated using scenes from different image databases.