10
OPTIMUM SCALE IN OBJECT-BASED
IMAGE ANALYSIS
JUNGHO IM, LINDI J. QUACKENBUSH, MANQI LI, AND FANG FANG
10.1 INTRODUCTION
Object-based image analysis (OBIA) or geographic OBIA (GEOBIA) has become
popular in the remote sensing literature since the introduction of high-spatialresolution (£1-m) satellite remote sensing data such as IKONOS and QuickBird.
Traditional pixel-based analysis often does not work well with such high-resolution
data due to high-frequency components and horizontal layover caused by off-nadir
look angles (Im et al., 2008b). Blaschke (2010) comprehensively reviewed over 820
articles dealing with object-based image analysis for remote sensing and provides an
overview of object-based approaches that are commonly used in the remote sensing
literature. The increased popularity of object-based image analysis represents a
significant trend in the remote sensing context as the number of publications in
the field continues to increase.
OBIA applications are broad and typically include image segmentation, edge
detection, feature extraction, classification, and change detection (Blaschke, 2010;
Stow, 2010). An image segmentation procedure is the first step commonly applied
in OBIA. This procedure divides the image into meaningful homogeneous and
nonintersecting regions based on spectral and/or spatial properties (Benz et al.,
2004). Such homogeneous regions can be hierarchically organized as image objects
(i.e., segments) (Blaschke, 2005). There are numerous algorithms developed for
image segmentation, which Cheng et al. (2001) groups into four types: (1)
histogram thresholding (Hofmann et al., 1998) that works for a single image
with several thresholds; (2) image feature space clustering (Hall et al., 1992), which
197
Scale Issues in Remote Sensing, First Edition. Edited by Qihao Weng.
Ó 2014 John Wiley & Sons, Inc. Published 2014 by John Wiley & Sons, Inc.
OPTIMUM SCALE IN OBJECT-BASED
IMAGE ANALYSIS
JUNGHO IM, LINDI J. QUACKENBUSH, MANQI LI, AND FANG FANG
10.1 INTRODUCTION
Object-based image analysis (OBIA) or geographic OBIA (GEOBIA) has become
popular in the remote sensing literature since the introduction of high-spatialresolution (£1-m) satellite remote sensing data such as IKONOS and QuickBird.
Traditional pixel-based analysis often does not work well with such high-resolution
data due to high-frequency components and horizontal layover caused by off-nadir
look angles (Im et al., 2008b). Blaschke (2010) comprehensively reviewed over 820
articles dealing with object-based image analysis for remote sensing and provides an
overview of object-based approaches that are commonly used in the remote sensing
literature. The increased popularity of object-based image analysis represents a
significant trend in the remote sensing context as the number of publications in
the field continues to increase.
OBIA applications are broad and typically include image segmentation, edge
detection, feature extraction, classification, and change detection (Blaschke, 2010;
Stow, 2010). An image segmentation procedure is the first step commonly applied
in OBIA. This procedure divides the image into meaningful homogeneous and
nonintersecting regions based on spectral and/or spatial properties (Benz et al.,
2004). Such homogeneous regions can be hierarchically organized as image objects
(i.e., segments) (Blaschke, 2005). There are numerous algorithms developed for
image segmentation, which Cheng et al. (2001) groups into four types: (1)
histogram thresholding (Hofmann et al., 1998) that works for a single image
with several thresholds; (2) image feature space clustering (Hall et al., 1992), which
197
Scale Issues in Remote Sensing, First Edition. Edited by Qihao Weng.
Ó 2014 John Wiley & Sons, Inc. Published 2014 by John Wiley & Sons, Inc.