Image-based building silhouette extraction

Abstract

Recently multiple methods and techniques have been developed to solve image segmentation problems in different domains. One of the problems is to detect building

from an common outdoor scene. Some methods have been developed to reach this goal

using different properties of buildings, such as vertical edges and colour features. Our project is based on colour, texture, oriented pattern and homogenous properties to distinguish buildings from background.

The project is to explore image segmentation of buildings and develop software to

realize this goal. The software is able to retrieve the silhouette of the main building from

a photo and to give the silhouette of the building. It is programmed using matlab. It

is inspired by multiple research papers, such as JSEG, Semantic Class Segmentation

using Oriented Pattern Analysis, and Maximum Similarity based Region Merging.