Automatic building extraction based on improved watershed segmentation, mutual information match and snake model

Authors:
Gang Li;Jinliang An;Youchuan Wan
Affiliations:
School of Remote Sensing and Information Engineering, Wuhan University, Wuhan city, Hubei Province 430079, China.;College of Information Technology, Henan Institute of Science and Technology, Xinxiang city, Henan Province 453003, China.;School of Remote Sensing and Information Engineering, Wuhan University, Wuhan city, Hubei Province 430079, China
Venue:
International Journal of Computer Applications in Technology
Year:
2012

Citing 8
Cited 1

A contour detection method: initialization and contour model

Pattern Recognition Letters
Analyzing the Mallat wavelet transform to delineate contour and textural features

Computer Vision and Image Understanding
Object Contour Extraction Using Adaptive B-Snake Model

Journal of Mathematical Imaging and Vision
Fast road network extraction in satellite images using mathematical morphology and Markov random fields

EURASIP Journal on Applied Signal Processing
A Framework for Image Segmentation Using Shape Models and Kernel Space Shape Priors

IEEE Transactions on Pattern Analysis and Machine Intelligence
Classified road detection from satellite images based on perceptual organization

International Journal of Remote Sensing
Adaptive fuzzy color segmentation with neural network for road detections

Engineering Applications of Artificial Intelligence
An adaptive greedy algorithm with application to nonlinear communications

IEEE Transactions on Signal Processing

Active AODE learning based on a novel sampling strategy and its application

International Journal of Computer Applications in Technology

Quantified Score

Hi-index	0.00

Visualization

Abstract

Automatic building extraction from high resolution remote sensing image has been a research hotspot. It is also a recognised difficulty because of the diversity and complexity of building structures. This paper proposes a new method for automatic building extraction based on improved watershed segmentation, mutual information match and improved snake model. Our method was divided into four stages, which were homogenous region generation, building template extraction, candidate building region selection and building boundary determination. Firstly, an improved watershed segmentation algorithm was proposed. The improvements included the adaptive de-noising method based on wavelet transformation and the marker extraction based on gradient flatness index. Secondly, building templates were extracted based on shape features and shape restrictions. Thirdly, candidate building regions were selected based on mutual information match. Finally, the traditional snake model was improved in two aspects: automatic determination of initial snake contour and new energy equation. According to the experiment results, our method can improve the accuracy of building extraction, and almost all the buildings are extracted correctly.