1. Brief Introduction
The release version of this software has been tested to run successfully over windows XP,
windows 7 operating systems. It is written by C++ language, compiled in Visual C++ 6.0.
OpenCV libraries are used to support some routines, such as region geometrical properties
computation, morphological operations, k-means clustering and SVM in the future.
The GDAL library is also used to read and write a host of image formats, *.tif, *.img, etc., and to
produce vector files, such as *.shp.
This software has realized the following methods
1.1.segmentation
Graph based segmentation [2] (GS);
Quad tree segmentation [1] (QT) (similar to that implemented in eCognition [6] and that as a
step in HSMR [7], but in a bottom-up fashion);
Hierarchical Clustering segmentation (HC) (i.e., Multiresolution segmentation [3]). It
adopted the basic data structures detailed in [5] and borrowed the idea of region adjacency
graph and nearest neighbor graph in [4]. There are two homogeneity criterion in the
implementation, the one used in ecognition and that used in ENVI EX feature extraction [5]
module. And two kinds of merging order for region pairs, one is to randomly select seed
regions for region growth in the manner of ecognition [6], the other is to select the region pair
with minimum homogeneity distance as done in [4] ;
The two-stage methods, GS-HC, QT-HC (described in [1] in detail)
Pyramid mean shift segmentation implemented in OpenCV 1.0
Mean shift segmentation [8] of Dorin Comaniciu (resembles EDISON [9])
Watershed segmentation [10] of Vincent and Soille
1.2.edge detection
SUSAN [11], Sobel, Canny edge detectors are incorporated.
1.3.classification
MRF classification initialized by K-means by ICM, Gibbs sampler, Metropolis etc.
- 1
- 2
前往页