Image analysis – Color image processing – Pattern recognition or classification using color
Reexamination Certificate
1999-08-30
2003-06-24
Au, Amelia M. (Department: 2623)
Image analysis
Color image processing
Pattern recognition or classification using color
C707S793000, C382S199000, C382S305000
Reexamination Certificate
active
06584221
ABSTRACT:
FIELD OF THE INVENTION
This invention relates generally to image processing, and more particularly to indexing, searching, and retrieving images by queries on image content.
BACKGROUND OF THE INVENTION
With the proliferation of multi-media, the world-wide web, and digital imaging, there now exists a demand for image management tools, most importantly tools for indexing, searching and retrieving images. This is commonly referred to as “query-by-image-content” (QBIC) or “content-based image retrieval” (CBIR). Existing systems often make use of global attributes such as overall color and texture distributions which ignore the actual composition of the image in terms of internal structures.
Most of the current content-based image retrieval systems rely on global image characteristics such as color and texture histograms, e.g., see Altavista's “Photofinder.” While these simple global descriptors are fast and often do succeed in partially capturing the essence of the user's query, global descriptors often fail due to the lack of higher-level knowledge about what exactly was of interest to the user in the query image, i.e., user-defined content. Recently, there has been a gradual shift towards spatially-encoded image representations. Spatially-encoded representations range widely from fixed image partitioning, as in the “ImageRover,” to highly local characterizations like the “color correlograms,” please see Sclaroff et al. in “Imagerover: A content-based image browser for the world wide web,” Proc. IEEE Workshop on Content-Based Access of Image and Video Libraries, June 1997, and Huang et al. in “Image indexing using color correlograms, “Proc. IEEE Conf. on Computer Vision and Pattern Recognition, 1997.
Somewhere in between these two extremes, one can find various techniques which deal with “regions” or “blobs” in the images. For example, “configural templates” specify a class of images, e.g., snow-capped mountain scenes, by means of photometric and geometric constraints. on pre-defined image regions, see Lipson et al. in “Configuration based scene classification and image indexing,” Proc. IEEE Conf. on Computer Vision and Pattern Recognition, 1997. Other techniques use automatic blob segmentation and description, see Carson et al. in “Region-based image querying,” Proc. IEEE Workshop on Content-Based Access of Image and Video Libraries, June 1997, and Howe in “Percentile blobs for image similarity,” Proc. IEEE Workshop on Content-Based Access of Image and Video Libraries, June 1998. Most of these systems require pre-segmented regions.
SUMMARY OF THE INVENTION
The invention provides a method for representing an image for the purpose of indexing, searching, and retrieving images in a multi-media database. The method allows a user to specify “content” to be searched as salient “regions-of-interest” or ROIs. The user also specifies the importance of the spatial relationships of ROIs. The method yields acceptable retrievals that are at least equal to global-based searches, and provides an intuitive interface that is more in tune with the user's notion of “content,” thus providing a more powerful image retrieval tool.
More particularly, the invention provides a method for representing an image in an image retrieval database. The method first separates and filters the image to extract texture features. The colors are readily obtained from the pixel values themselves. The color and texture features are partitioned into a plurality of blocks, each block is 16×16 pixels. A joint distribution of the color features and a joint distribution of the texture features are estimated for each block. The estimated joint distributions, expressed as histograms, are stored in the database with the image.
In one aspect, the color features include three color coordinates LUV, and the texture features include three edge measurements: edge magnitude, a rotation invariant Laplacian, and an edge orientation. The edge magnitude is log(G
2
x
+G
2
y
), the rotation invariant Laplacian is G
xx
+G
yy
, and the edge orientation is arg(G
x
, G
y
), where G
x
, G
y
and G
xx
, G
yy
are respectively the first and second derivatives of a Gaussian filter with specified scale &sgr; with two parameters used for the scale &sgr;. In a query image, one or more regions of interest can be specified with a user interface, and joint distributions can be estimated for blocks corresponding to the regions of interest at as above. The query joint distributions can be used to index the database to find matching images. The matching images can be rank ordered. During the matching, the spatial relationship of the regions of interest can be considered in the matching.
REFERENCES:
patent: 5675506 (1997-10-01), Savic
patent: 5751286 (1998-05-01), Barber et al.
patent: 5767978 (1998-06-01), Revankar et al.
patent: 5987456 (1999-11-01), Ravela et al.
patent: 5995115 (1999-11-01), Dickie
patent: 6253201 (2001-06-01), Abdel-Mottaleb et al.
patent: 6363381 (2002-03-01), Lee et al.
patent: 2001/0053249 (2001-12-01), Krishnamachari
Smith et al. “Local Color and Texture Extraction and Spatial Query”. Image Processing, 1996. IEEE Proceedings., International Conference on, vol.: 3, 1996. Pp.: 1011-1014 vol. 3.*
Jähne. “Practical Handbook on image Processing for Scientific Applications”. CRC Press, 1997, p. 389-445.*
C. Carlson et al.; “Region-Based Image Querying”; In Proc. IEEE Workshop on Content-Based Access of image and Video Libraries; Jun., 1997.
N. Howe; “Percentile Blobs for Image Similarity”; IEEE Workshop on Content-Based Access of Image and Video Libraries; Jun., 1998.
J. Huang; “Image Indexing Using Color Correlograms”; In Proc. IEEE Conf. on Computer Vision and Pattern Recognition, 1997.
P. Lipson et al.; “Configuration Based Scene Classification and Image Indexing”; In Proc. IEEE Conf. on Computer Vision and Pattern Recognition, 1997.
S. Sclaroff; “ImageRover: A Content-Based Image Browser for the World Wide Web”; In Proc. IEEE Workshop on Content-Based Access of Image and Video Libraries; Jun., 1997.
Biermann Henning
Margaritis Dimitris
Moghaddam Baback
Au Amelia M.
Brinkman Dirk
Curtin Andrew J.
LaRose Colin
Mitsubishi Electric Research Laboratories Inc.
LandOfFree
Method for image retrieval with multiple regions of interest does not yet have a rating. At this time, there are no reviews or comments for this patent.
If you have personal experience with Method for image retrieval with multiple regions of interest, we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Method for image retrieval with multiple regions of interest will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-3090962