Invariant texture matching method for image retrieval

Image analysis – Pattern recognition – Feature extraction

Reexamination Certificate

Rate now

  [ 0.00 ] – not rated yet Voters 0   Comments 0

Details

C382S181000, C382S191000, C382S276000, C382S305000, C707S793000, C707S793000, C707S793000

Reexamination Certificate

active

06192150

ABSTRACT:

BACKGROUND OF THE INVENTION
1. Field of the Invention
This invention relates to a method and an apparatus for matching a sample texture to those contained in a large collection of images; and, more particularly, to a method and an apparatus which matches texture patterns independently of the intensities, scales, and orientations of the patterns.
2. Description of Prior Art
Texture relates to a human's perception of visual characteristics such as smoothness, coarseness, and regularity of various materials. Many objects, such as brick walls, bushes, roof tiles, and fabric, can be recognized or recalled based on the distinctive texture patterns the objects contain. Therefore, texture is a very important visual cue for retrieving user-required images from large image databases.
In general, a texture matching technique for image retrieval compares a query texture image with those contained in a plurality of database images, and retrieves those database images which contain one or more texture patterns that are similar to the query texture. There are generally three conventional approaches for performing texture matching or classification, namely, the structural approach, the statistical approach, and the spectral approach.
The structural approach characterizes a texture by determining the spatial arrangement of visual primitives such as line segments, line ends, and blobs of pixels. Although there is psychological evidence supporting such a structural approach, it is difficult from a computational standpoint to determine the visual primitives and their spatial arrangements.
The statistical approach characterizes a texture in terms of numerical attributes such as local statistics and simultaneous regression. These attributes describe the statistical distribution of intensity values around a pixel. Although these numerical attributes are generally easy to compute, they do not provide a simple means of performing scale- and orientation-invariant texture matching.
The spectral approach filters texture images using a set of filters, and uses the filtered outputs as features for texture classification. Gabor filters are most commonly used for this purpose. Gabor filters can extract frequency and orientation information from the texture images. The Gabor filtering approach has been used in W. Y. Ma and B. S. Manjunath, “Texture Features and Learning Similarity,”
Proceedings of IEEE Conference on Computer Vision and Pattern Recognition,
pages 1160-1169, 1996; German Patent DE 4406020; and Japanese Patent JP 09185713.
Except for the work of the present Inventors, the above-mentioned prior techniques, however, do not provide a texture matching method that is invariant to both scale and orientation. Furthermore, such existing texture matching methods assume that each image contains only a single uniform texture pattern.
SUMMARY OF THE INVENTION
It is the primary object of the present invention to provide an intensity-, scale-, and orientation-invariant texture matching apparatus and method that computes the similarity between a query texture and a plurality of images, which may contain one or more texture patterns of varying intensities, scales, and orientations, and produces a ranking of the plurality of images according to the computed similarity.
The aforementioned and other objects of the present application are achieved by providing a texture matching apparatus comprising:
a texture feature extractor for identifying a plurality of regions in a plurality of texture images, each region containing a texture pattern, extracting an N-component texture feature vector for each region, and normalizing the extracted N-component texture feature vectors;
a feature transformer for transforming the normalized N-component feature vectors into D-dimensional vector points in a D-dimensional vector space that is intensity-, scale- and orientation-invariant; and
an image ranker that ranks the plurality of texture images according to their similarity to a query texture measured in the D-dimensional invariant texture space.
The aforementioned and other objects are further achieved by providing a texture matching method comprising:
identifying from a plurality of texture images a plurality of regions, each containing a texture pattern;
extracting an N-component texture feature vector for each region;
normalizing the extracted N-component texture feature vectors;
transforming the N-component feature vectors into D-dimensional vector points in a D-dimensional texture space that is invariant to the intensities, scales, and orientations of the texture patterns; and
ranking the plurality of images according to their similarity to the query texture measured in the D-dimensional invariant texture space.
In one embodiment of the present invention, the texture feature extractor includes a plurality of Gabor filters with different center spatial frequencies and orientations. The preferred feature transformer includes a plurality of masks that are applied to the feature vectors to measure texture characteristics along the plurality of dimensions of the invariant texture space.
The preferred image ranker computes the similarity between two texture patterns according to the difference between their corresponding vector points in the invariant space and the difference between their feature vectors.
Further scope and applicability of the present invention will become apparent from the detailed description given hereinafter. However, it should be understood that the detailed description and specific examples, while indicating preferred embodiments of the invention, are given by way of illustration only, since various changes and modifications within the spirit and scope of the invention will become apparent to those skilled in the art from the detailed description.


REFERENCES:
patent: 5579471 (1996-11-01), Barber et al.
patent: 5751286 (1998-05-01), Barber et al.
patent: 5913205 (1999-06-01), Jain et al.
patent: 5915250 (1999-06-01), Jain et al.
patent: 4406020C1 (1995-06-01), None
patent: 63-092445 (1988-04-01), None
patent: 9185713 (1997-07-01), None
Manjunath, et al. “texture features for browsing and retrieval of image data”, IEEE, pp. 837-842, Aug. 1996.
Remias, et al. “block-oriented image decomposition and retrieval in image database systems”, IEEE, pp. 85-92, May 1996.
Lu, et al. “texture segmentation by clustering of gabor feature vectors”, IEEE, pp. 683-688, Jan., 1991.
Dunn “an invariant pattern representation based on nonuniform sampling in the human visual system”, IEEE. pp. 600-603, 1994.
Teuner, et al. “orientation-and scale-invariant recognition of textures in multi-object scenes”, IEEE, pp. 174-177, Jul. 1997.
Xie, et al. “wavelet multiscale representation and morphological filtering for texture segmentation”, IEEE, pp. 1-8, 1993.
A structural Analyzer for a Class of Textures, Saburo Tsuji et al., Computer Graphics and Image Processing (1973) 2, 216-231.
The QBIC Project: Querying Images By Content Using Color, Texture, and Shape, W. Niblack et al., SPIE vol. 1908 (1993) pp. 173-181.
Detection of Homogeneous Regions By Structural Analysis, Fumiaki Tomita et al., Department of Control Engineering, Osaka University, pp. 564-571.
Content-based Retrieval of Segmented Images, Tat-Seng Chua et al., Department of Information Systems and Computer Science, pp. 211-218.
Josef Bigün et al., IEEE Transactions on Pattern Analysis And Machine Intelligence, vol. 16, No. 1, Jan. 1994 pp. 80-87.
Alan Conrad Bovik et al., IEEE Transactions on Pattern Analysis And Machine Intelligence, vol. 12, No. 1, Jan. 1990, pp. 55-73.
Dennis Dunn et al., IEEE Transactions on Pattern Analysis And Machine Intelligence, vol. 16, No. 2, Feb. 1994, pp. 130-149.
Seow Yong Lai et al., Invariant Texture Matching for Content-Based Image Retrieval, pp. 53-68.
Fang Liu et al., IEEE Transactions on Pattern Analysis And Machine Intelligence, vol. 18, No. 17, Jul. 1996 pp. 722-733.
W. Y. Ma et al., IEEE 1996, pp. 425-430.
Jianchang Mao et al., Pattern Recognition, vol. 25, No. 2, pp. 173-188.
A. Ravishankar Rao, IEEE Conference on Visualizat

LandOfFree

Say what you really think

Search LandOfFree.com for the USA inventors and patents. Rate them and share your experience with other people.

Rating

Invariant texture matching method for image retrieval does not yet have a rating. At this time, there are no reviews or comments for this patent.

If you have personal experience with Invariant texture matching method for image retrieval, we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Invariant texture matching method for image retrieval will most certainly appreciate the feedback.

Rate now

     

Profile ID: LFUS-PAI-O-2592336

  Search
All data on this website is collected from public sources. Our data reflects the most accurate information available at the time of publication.