Image analysis – Pattern recognition – Feature extraction
Reexamination Certificate
1998-08-20
2002-04-30
Au, Amelia M. (Department: 2623)
Image analysis
Pattern recognition
Feature extraction
C382S218000, C358S521000
Reexamination Certificate
active
06381365
ABSTRACT:
This application is based on patent application Nos. 9-226101 and 9-235979 filed in Japan, the content of which is incorporated hereinto by reference.
BACKGROUND OF THE INVENTION
1. Field of the Invention
The present invention relates to an image processing apparatus and image processing method for extracting features from an image, and relates more specifically to an apparatus and method for extracting the textural features of an image.
2. Description of the Related Art
Image databases typically contain various attributes and search information for each image in the database. Attributes typically include, for example, the original image production date, revision date, file name, and file format. Search information typically includes, in addition to the attributes, keywords and feature information such as shapes, colors, and even audio information. The attributes are often attached automatically to a new image when the image is registered into the database. Features such as texture, color, and edge information can also be automatically extracted from the image and added to the database. The problem with the feature extraction is the processing time required to extract the features from the images. This is a particular problem with textural features because of the number of operations required.
There has therefore been a need for a method of efficiently extracting image features, and particularly textural features, from image information.
Japanese Patent Laid-Open Publication No. 5-28266 teaches a method for extracting texture features from an image by segmenting an image into square blocks and then extracting feature information from each block. This method, however, divides the entire image into the blocks from which features are then extracted, and thus processes the entire image. The taught method is thus not an efficient means of extracting features from images having a uniform texture pattern throughout. This method also does not extract structural information about texture patterns.
Also the method of extracting information about texture structure is disclosed in, for example, “Basics of image recognition II: feature extraction, edge detection, and texture analysis (in Japanese, by Mori Shunji et al.; Ohmsha)”. This teaches a method for extracting the basic pattern and arrangement of pattern elements creating a particular image texture using a rigid processing method, but the process is complicated and is not practical in terms of the required processing speed.
Image feature extraction is normally applied to image data after image normalization, gray scale conversion, and digitizing have been completed. The threshold values used for digitizing images are normally set to a median value at which pixel “0” and pixel “1” appear with same probabilities in pixel distribution, or to the average brightness level of the image. Because this often results in the threshold value being set to the pixel value of image areas having a gradual tonal change and occupying a large area of the image, the final digital image tends to look unnatural when the image has a substantially uniform background such as blue sky or other solid color image. The problem is, more specifically, that when digitizing images with sky or basically solid color background, parts of the background are converted to pixel “0” and other parts are converted to pixel “1” even though there is no change in the background, resulting in an unnatural appearance. When the resulting digital image data is then used for feature extraction, image features cannot be appropriately extracted.
There is therefore a need for a means and method whereby texture features can be extracted from an image efficiently and accurately.
There is a further need for a means and method whereby image searching can be performed with good precision by extracting the basic texture pattern of an image as a feature of the image, and comparing images using these detected basic texture patterns.
SUMMARY OF THE INVENTION
An object of the present invention is therefore to provide an image data processing apparatus and method for efficiently and accurately extracting texture features from images.
A further object of the invention is to provide an image data processing apparatus and method for searching image data by extracting feature information relating to basic image texture patterns, and comparing extracted patterns to determine image content.
In an aspect of this invention, an image data processing apparatus according to an exemplary embodiment of the present invention addresses particularly detecting texture features from image data by means of a texture extraction controller, a memory and a comparison controller.
The texture extraction controller obtains feature information relating to said image texture based on the image data. The memory stores image data and the associated feature information extracted from the image data by the texture extraction controller with a defined correlation between the stored image data and feature information so that the comparison controller can seek images in memory by comparing images based on the feature information stored to memory.
In another aspect of this invention, an image data processing apparatus processes image data for images containing a texture formed by a basic pattern repeating at a specific period by means of a detection controller, a memory and a comparison controller. The detection controller detects feature information relating to the basic pattern of the image texture and the arrangement of basic pattern elements based on source image data. The memory stores image data and the associated feature information detected by the detection controller with a defined correlation between the stored image data and feature information so that the comparison controller can compare images based on the stored feature information.
In another aspect of this invention, an image data processing apparatus processes the image data for images containing a texture pattern formed by a basic pattern repeating at a specific period, and comprises an image database, an input means, and a comparison and searching controller. The image database stores image data for a plurality of images, and feature information describing a basic pattern and period of a texture in an associated stored image. The relationship between image data and associated feature information is defined such that a particular image can be found by referencing the associated feature information. The input means enables a user to select an image to be used as a search key whereby the image data processing apparatus seeks a desired image data from the image database. The comparison and search controller seeks a desired image from the image database by sequentially comparing feature information for the image selected as a search key with feature information stored in the image database, and determines any similarity between the search key and compared image.
In further aspect of this invention, an image data processing apparatus processes image data for a pixel image wherein a plurality of pixels are arranged in a matrix pattern and the pixel image has a texture pattern. To process the image data, this image data processing apparatus uses an extraction controller for extracting feature information descriptive of an image texture using pixel data for pixels located in a scanning band oriented in a specific direction through the image. The scanning band can be a line or a band of a specific width.
REFERENCES:
patent: 5085517 (1992-02-01), Chadwick et al.
patent: 5513275 (1996-04-01), Khalaj et al.
patent: 5533148 (1996-07-01), Sayah et al.
patent: 5574886 (1996-11-01), Koike et al.
patent: 5668897 (1997-09-01), Stolfo
patent: 5734747 (1998-03-01), Vaidyanathan
patent: 5859930 (1999-01-01), Chase
patent: 5887080 (1999-03-01), Tsubusaki et al.
patent: 6222932 (2001-04-01), Rao et al.
patent: 60-157681 (1985-08-01), None
patent: 5-28266 (1993-02-01), None
patent: 5-274372 (1993-10-01), None
patent: 6-274193 (1994-09-01), None
patent: 6-295318 (1994-10-01), None
pa
Au Amelia M.
Burns Doane , Swecker, Mathis LLP
Dastouri Mehrdad
Minolta Co. , Ltd.
LandOfFree
Image data processing apparatus and image data processing... does not yet have a rating. At this time, there are no reviews or comments for this patent.
If you have personal experience with Image data processing apparatus and image data processing..., we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Image data processing apparatus and image data processing... will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-2881596