Image retrieval system and method using image histogram

Details Image retrieval system and method using image histogram Image retrieval system and method using image histogram

2000-02-04

2003-09-16

Au, Amelia M. (Department: 2721)

Image analysis

Histogram processing

C382S305000

Reexamination Certificate

active

06621926

ABSTRACT:

TECHNICAL FIELD
The present invention relates to an image retrieval system and method using an image histogram, and, in particular, to an image retrieval system and method of using an image histogram, for determining central points and dispersion values as well as quantity information of color about respective histogram bins, thereby using these as mapping information for image retrieval.
BACKGROUND OF THE INVENTION
An image database can be classified into three groups according to retrieval methods, i.e., text-based, content-based, and semantic-based database, and features used in the retrieval of content-based database are shape, texture, and color. In this case, color has easier to draw characteristics and more applicable to a user's query-by-example than shape and texture.
Conventional color indexing systems usually utilize only a bin which is a specific value of a color histogram. Here, the method of measuring similarity using the difference between two color histograms often results in the difference from the measurement result of the perceptual similarity of an actual person.
This is because color feature values of the histogram, i.e., values of each bin show global feature information, and it is difficult to retrieve an image having correctly requested contents with only the global feature information. That is, a global feature is advantageous for not being affected by rotation of the image or a slight change of position, but has a drawback of not containing any spatial information. Because of such characteristics of a global feature that does not contain spatial information, when retrieving with only color information, a false positive error in the retrieval result can occur.
Since it is very difficult to draw spatial information from an image, conventional methods, which usually divide an image into sub-regions, and obtain global features for the respective sub-regions, have been proposed. Such a method is referred to the color layout. The simplest color layout indexing method divides an image into blocks of the same size and draws color features from the respective blocks, but this method is not suitable for an image having great color variance. Another method performs a segmentation of an image and obtains features of each segment.
It is difficult to implement a full automation system since it is difficult to divide objects of an image. In the real system, a dividing work that users semi-automatically or manually determine objects is used. On the other hand, in recent years, as a new color layout indexing method, methods using the wavelet transform are proposed.
As mentioned above, conventional methods perform retrieval by obtaining global color information and dividing a region in a color histogram and drawing spatial information from the respective sub-regions, and then synthesizing said two results. Such method is not effective because of a problem of synthesizing said two results since there are no relationships between two obtained information, and because it is necessary to draw two different features twice.
SUMMARY OF THE INVENTION
The disclosed embodiments of the present invention provide an image retrieval system and method using an image histogram for finding central points and dispersion values as well as quantity information of color about respective histogram bins, thereby using these as mapping information for image retrieval.
In order to achieve the foregoing, the embodiments of the present invention provide an index information generator of an image retrieval system using an image histogram, wherein said index information generator comprises operation means for computing histogram image bins when an image is inputted, and accumulating x, y, x
2
, and y
2
, to compute central points and dispersion values; and the first normalization means for dividing the respective computed central points and dispersion values by the size of the whole image.
In addition, the embodiments of the present invention provide a query image retriever of an image retrieval system using an image histogram, wherein said query image retriever comprises generation means for drawing feature vectors to generate values of a model to be retrieved, operation means for computing a difference between said generated values of the model and the number, count, central point, and dispersion values of corresponding stored histogram bins, and means of specifying similarity for specifying the similarity value of an image by using said computed values.
Furthermore, the disclosed embodiments provide a method for image retrieval using an image histogram that includes the following steps. A first step is when an image is inputted, computing histogram image bins and computing central points and dispersion values by accumulating x, y, x
2
, and y
2
. A second step is normalizing through dividing respective central points and dispersion values computed in the first step by the size of whole image, and storing it. A third step is when a query image is inputted, drawing feature vectors and generating values of model to be retrieved, and then computing differences between said generated values of model and the number, count, and central points and dispersion values of the corresponding bins of the image histogram. A fourth step is specifying the similarity value of image using the values computed in the third step.
Also, the disclosed embodiments provide a storage medium containing a program that executes steps, including the following steps. A first step is when an image is inputted, converting the image into color coordinate system, and normalizing it to reduce the feature of the converted values. A second step is computing histogram color bins from the normalized values in the first step, and accumulating x, y, x
2
, and y
2
, thereby computing central points and dispersion values. A third step is normalizing the respective computed central points and dispersion values by dividing with the size of whole image, and storing it. A fourth step is when a query image is inputted, generating a value of model to be retrieved by drawing a feature vector, and then computing the difference between the generated value of model and the number, count of color values, and central points and dispersion values of bins corresponding to the stored data in the third step. And a fifth step is specifying the similarity values of an image using the computed values in the fourth step.


REFERENCES:
patent: 6163622 (2000-12-01), Abdel-Mottaleb et al.
patent: 00 53072 (1999-07-01), None
Smith et al. “Single Color Extraction and Image Query”. Proc. of International Conference on Image Processing, V. 3, p. 528-531, 1995.*
Zhang et al. “Comparison and Improvement of Color-Based Image Retrieval Techniques”. SPIE V. 3312, p. 371-382, 1997.*
Rickman et al. “Content-Based Image Retrieval Using Colour Tuple Histograms”. SPIE V. 2670, p. 2-7, 1996.*
John R. Smith et al., “Tools and Techniques for Color Image Retrieval”, IS&T/SPIE Proceedings, vol. 2670, Storage & Retrieval for Image and Video Databases IV, pp. 1-12, Feb. 1, 1996.
W. Niblack et al., “The QBIC Project: Querying Images by Content Using Color, Texture, and Shape”, IBM Research Division, SPIE, vol. 1908, pp. 173-187, 1993.
James Ze Wang et al., “Wavelet-Based Image Indexing Techniques with Partial Sketch Retrieval Capability”, Proceeding on the Fourth Forum on Research and Technology Advances in Digital Libraries, pp. 1-12, May 1997.

Affiliated with

Also associated with

Rating

Image retrieval system and method using image histogram 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 retrieval system and method using image histogram, we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Image retrieval system and method using image histogram will most certainly appreciate the feedback.

Rate now

Comments { 0 }

Profile ID: LFUS-PAI-O-3020379