Image analysis – Color image processing – Pattern recognition or classification using color
Reexamination Certificate
1999-01-29
2002-09-03
Johnson, Timothy M. (Department: 2623)
Image analysis
Color image processing
Pattern recognition or classification using color
C382S170000, C382S305000, C707S793000
Reexamination Certificate
active
06445818
ABSTRACT:
BACKGROUND OF THE INVENTION
1. Field of the Invention
The present invention relates to a color image processing method, and more particularly to a method for designating a local representative color value and automatically determining an optimal image searching algorithm on a color image.
2. Discussion of Related Art
In the field of color image analysis and processing, extensive researches for a method to accurately search images are actively being conducted. In response, commercial image searching apparatus and/or applications program are being developed to meet the demand for an improved image searching.
An image searching system based on content of the image is generally divided into a feature extraction module extracting characteristics of the images, an image searching module searching the images, and a module of image database.
To expedite the searching process, the characteristics of images are extracted as the images are input to the image database and the extracted characteristics are also stored in the image database. Subsequently, if a user requires an image search, a similarity between respective images are determined utilizing the characteristic information of a reference image and target images stored in the database. The target images are sorted according to the result of the comparison.
The most important information for use in the content based image searching is the color information. Accordingly, an effective performance of an image searching apparatus or applications program depends greatly on an accurate method for extracting the color information.
Generally, the number of different colors which can theoretically be expressed by a computer has been steadily increasing. However, the number of colors which can be displayed by the computer is limited by the available number of quantized colors.
In computers, a color is expressed utilizing the RGB color model based upon the three primary colors of red R, green G, and blue B. However, the RGB space is hardware oriented and a limitation exists in expressing the color changes such that the change can be visually sensed by the human eye. Thus, the hue, chroma, and brightness of the RGB space is often converted into a user oriented HSV color model based upon a hue H, saturation S and value V, then converted back to the RGB space by quantization. The CIE standards may be utilized to convert from the RGB space to the HSV color space.
Numerous amount of content based image searching apparatuses or applications programs utilizing the HSV space have been already proposed in the related art.
The characteristic information used for image searching are extracted from every pixel of an image and are mapped to color indexes. The weight of each color index on the image may be represented by a color histogram. Thus, the color histogram is an important data, indicating a color distribution of the image. Generally, there are two types of color histograms represented by n number of quantized colors, namely a global color histogram and a local color histogram.
In the local color histogram, an image is divided into n number of grids and a local histogram for each grid cell is determined. A representative index color of each local histogram is defined as a local representative color of the corresponding grid.
Thus, the global color histogram indicates a color distribution in an overall image and represents the total distribution of colors for respective pixels of the overall image, and the local color histogram indicates a distribution in a local grids and represents the total distribution of the color on a specific region of the image.
To build a color histogram, the RGB values in each pixel of an input image is converted a user oriented color space, and the converted RGB values are mapped to one of n number of colors according to its quantized color area. Based upon the mapped color, the global color histogram and the local color histogram are constructed for all image pixels. Subsequently, the histograms of a reference image and target images are compared, and arranged in the order of highest similarity. Such order becomes a sequential array of most similar images by the color information.
In the content based image searching system, accurate color values of the image must be obtained to perform an effective analysis.
FIGS. 1A through 1E
show an example method for designating a local representative color value on a color image.
FIG. 1A
shows an image to be analyzed,
FIG. 1B
indicates a color histogram for an overall image,
FIG. 1C
shows the image by block or grid state,
FIG. 1D
sets forth a local histogram for each grid; and
FIG. 1E
shows the designated local representative color value for each grid.
Particularly, the color characteristic of the image is represented by the global color histogram as shown in FIG.
1
B. The image has also been partitioned into block areas with constant sizes for a local consideration of the image. Accordingly, the color characteristic of each block areas is represented by the local color histogram as shown in FIG.
1
D.
If only the global color histogram as shown in
FIG. 1B
is utilized in the image search, local contents of the image would not be considered. However, if the local color histogram as shown in
FIG. 1D
is also utilized to take into consideration the local contents, a memory with a large storage capacity would be required for storing each local color histogram data. Also, it would be difficult to easily represent the color characteristic of the image.
Numerous search algorithms according to the image characteristics have been conventionally utilized in the content base image searching system. The search algorithm for the image searching system is typically based upon a local representative color, a major color region (MCR), and a global color histogram.
The local representative color indicates a representative color value for each grid of a divided image. If there is no color representable in a grid, a “Don't Care” symbol would be represented. The major color region indicates a position on which a major color of the image is represented. For example, the major color grid may be represented as a minimum square grid. The major color here means a color which has been distributed more than a specific threshold.
As discussed above, in the content base image searching system of the related art, numerous algorithm may be used in considering several different search characteristics. Moreover, in searching images, a user manually controls the weight of the algorithm through a user interface. However, it is very difficult for a user to directly control the image searching algorithm during an image search.
Especially, there may be a significant differences in visual angle or sensual level of a person resulting in different contents stored in the database. Furthermore, for a user who is not a highly trained expert group, manually controlling the weight of numerous search algorithm is very difficult.
Finally, the differences in the visual angle or sensual level of persons makes it difficult to determine the weight of the optimum search algorithm while maintaining a high speed image searching system.
SUMMARY OF THE INVENTION
Accordingly, an object of the present invention is to solve at least the problems and disadvantages of the related art.
An object of the present invention is to provide a method for designating a local representative color value on a color image.
Another object of the present invention is to provide a method for an automatic selection of the most proper search algorithm in an image searching system.
Still another object of the present invention is to provide a method for an automatic selection of a search algorithm in a content base image searching system utilizing a characteristic weight.
Additional object of the present invention is to provide a method for auto-determining the detection algorithm in an image detector capable of furnishing linearly the weight according to the number of color blocks of the local grid.
Additional advantages, objects, and featu
Kim Hyeon Jun
Lee Jin Soo
Fleshner & Kim LLP
Johnson Timothy M.
LG Electronics Inc.
LandOfFree
Automatically determining an optimal content image search... does not yet have a rating. At this time, there are no reviews or comments for this patent.
If you have personal experience with Automatically determining an optimal content image search..., we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Automatically determining an optimal content image search... will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-2875354