A rotation- and flip-invariant algorithm for representing spatial continuity information of geographic images in content-based image retrieval
Introduction
Content based image retrieval (CBIR) has gained increasing attention from GIS scientists (Manjunath and Ma, 1996; Bruns and Egenhofer, 1996; Sheikholeslami et al., 1999; Agouris et al., 1999; Stefanidis et al., 2002). Originating from the computer vision and database community (Flickner et al., 1995; Pentland et al., 1996; Smith and Chang, 1996), CBIR attempts objectively and efficiently to retrieve targeted images or image regions from a large-volume image database based on the content similarity between a query icon (or sample image) and database images. What distinguishes it from conventional image retrieval is that in CBIR, image content is represented using numeric measurements of visual properties such as color, texture and shape (Flickner et al., 1995; Pass and Zabih, 1996; Rui et al., 1999).
While the interest in CBIR for analysis of geographic-image databases are rising, more attention needs to be on the uniqueness of geographic images and appropriate visual content representations. Recent CBIR research in geographic applications mostly focused on developing general CBIR approaches and applying them in retrieving texture-rich geographic entity types, such as forest, farmland, grasses, and others (Manjunath and Ma, 1996; Sheikholeslami et al., 1999). Man-made objects, such as factories, shopping centers and roads, in contrast, are much less studied, despite the fact that retrieving these man-made objects has important and broad applications, especially for urban areas. These man-made objects are discrete, but embedded in a complex background, and may not have rich texture patterns: thus retrieving these man-made objects presents a challenge conceptually as well as methodologically. Some relevant researches along this direction are reported recently (Bruns and Egenhofer, 1996; Agouris et al., 1999; Stefanidis et al., 2002), and they are based on object layers and assume that appropriate delineation of such objects has been completed either from geographic images or other sources. However, for images with complex scene structures like geographic images, automatic object delineation is still not an easy task, if not impossible. This paper presents an algorithm effectively to represent geographic-image content for CBIR, and its representation is directly derived from raw images. The algorithm is based on a well-recognized property of geographic space—spatial continuity.
In the following sections, a short discussion is first devoted on the process of general CBIR and its relationships with image classification. We then briefly review the concept of spatial continuity and its relevance in geographic-image study. Next, the new spatial continuity-based algorithm for CBIR is presented and experimental results are reported. Finally, we draw brief conclusions and make some discussions
Section snippets
CBIR process and image classification
The general CBIR process is illustrated in Fig. 1 and at least three components are deemed important: image segmentation, design of a numeric index vector (or visual property representation), and similarity measurement. First an image may need to be segmented into regions with homogeneous semantics. To represent its semantic content, generally a numeric index vector will be derived, with each element being a measurement of a visual property. Then, similarity between two image regions is
Spatial continuity and variogram
Spatial continuity can be defined as “the propensity for nearby locations to influence each other and to possess similar attributes” (Goodchild, 1992). It is a distinctive property of earth phenomena and exists in most geographic data sets, including geographic images. When there is no difference along different directions, it is described as isotropy. However, it is not uncommon that spatial continuity is not the same along different directions. This is termed spatial continuity anisotropy and
Design of the numeric index vector
As discussed in Section 2, the core for the new algorithm is to design the numeric index vector to represent image content. The basic idea behind the new design is to use a set of semi-variances at chosen lags and selected directions, so that they can represent the essence of several key directional variograms, and subsequently the spatial continuity and continuity anisotropy in an image. By reordering semi-variances based on directional difference of spatial continuity, it is hoped the
Test the anisotropy alignment capacity of the new algorithm
The first test aims to evaluate the performance of the algorithm for spatial continuity anisotropy alignment. Simple “L” images with distinct anisotropy are used (Fig. 8).
In the original image (Fig. 8), the major direction of spatial continuity is apparently north–south (90–270°). The ‘L’ shape is flipped and rotated in 90°, 180°, and 270° to create four additional images. Then we verify whether the new algorithm could align the images according to the spatial continuity of the ‘L’ shape. In
Conclusions and discussions
In this paper, we introduce a rotation- and flip-invariant algorithm for content-based geographic image retrieval, based on spatial continuity information. The construction units of image visual property representation are a set of semi-variances calculated at selected lag distances and along chosen directions. Lags are selected to capture the basic forms of variograms of image regions representing typical geographic entity types. Thus it can convey the continuity information in geographic
Acknowledgments
The author greatly appreciates the graduate assistantship from the National Center of Geographic Information Analysis and Geography Department in SUNY at Buffalo. The comments from reviewers and editors help improve the manuscript.
References (27)
- et al.
An environment for content-based image retrieval from large digital databases
ISPRS Journal of Photogrammetry and Remote Sensing
(1999) - et al.
Geostatistical classification for remote sensingan introduction
Computers & Geosciences
(2000) The semi-variogram in remote sensingan introduction
Remote Sensing of Environment
(1988)- et al.
Image retrievalcurrent techniques, promising directions, and open issues
Journal of Visual Communication and Image Representation
(1999) - et al.
Characterizing the spatial structure of vegetation communities in the Mojave Desert using geostatistical techniques
Computers & Geosciences
(2000) - et al.
Content-based image database retrieval using variance of gray level spatial dependencies
Lecture Notes in Computer Science
(1998) - Anderson, J.R., Hardy, E.E., Roach, J.T., Witmer, R.E., 1976. A land use and land cover classification system for use...
Multiscale nature of spatial data in scaling up environmental models
- Bruns, H.T., Egenhofer, M.J., 1996. Similarity of spatial scenes. In: Proceedings of tthe Seventh International...
- et al.
Principles of Geographical Information Systems
(1998)
Query by image and video contentthe QBIC system
IEEE Computers
Geographical information science
International Journal of Geographical Information Systems
Cited by (7)
Wavelet optimization for content-based image retrieval in medical databases
2010, Medical Image AnalysisObject-based target search using remotely sensed data: A case study in detecting invasive exotic Australian Pine in south Florida
2008, ISPRS Journal of Photogrammetry and Remote SensingCitation Excerpt :The general GIR process is illustrated in Fig. 1. Three components are especially important: image segmentation, content representation, and similarity measurement (Xie, 2004). An image is first segmented into individual regions (image objects), which hopefully contain homogeneous semantic content.
SAR image retrieval based on semi-variances and moment invariants
2010, Guangdian Gongcheng/Opto-Electronic EngineeringUsing spatial continuity and discontinuity information to retrieve geographic entities
2009, Geocarto InternationalRetrieval of remote sensing image based on combining spatial relation with texture feature
2008, Journal of Information and Computational ScienceIncorporating multiple index vectors and feedback to improve urban objects retrieval
2005, GIScience and Remote Sensing