Image analysis – Image segmentation
Reexamination Certificate
1998-09-30
2003-04-29
Wu, Jingge (Department: 2623)
Image analysis
Image segmentation
C382S299000, C358S462000, C358S464000
Reexamination Certificate
active
06556711
ABSTRACT:
BACKGROUND OF THE INVENTION
1. Field of the Invention
This invention relates to an image processing apparatus and method. More particularly, the invention relates to an image processing method and apparatus for extracting image segments having different characteristics from an input color image and judging each of the characteristics of these image segments.
2. Description of the Related Art
Apparatuses for inputting and outputting color images have begun to find wider use in the field of office automation. This has been accompanied by greater opportunities for handling color images obtained by using an input device such as a scanner to read an original of a color image the greater part of which is occupied by a black-and-white portion or a portion wherein the number of colors is limited. Such a color image shall be referred to as a “document image”. A feature of a document image is that, unlike other images, a document image includes a mixture of image segments (such as character image segments and photographic image segments) having different properties. There are cases in which an image having good picture quality is not obtained even if use is made of processing techniques utilized in other fields, such as a technique for zooming or a technique for reducing the quantity of data. Further, there is increasing demand for re-utilizing digital data obtained by structuring the color image of an original. More specifically, this entails inputting, into a computer, characters and photographs from the original in the form of character codes and image data, respectively, whereby the inputted information can be re-utilized in, say, desktop publishing (DTP).
Next, a technique for subjecting an image to zoom processing, a technique for reducing the quantity of data and a technique for image segmentation (inclusive of a technique for structuring a document image) will be described.
[Zoom Processing]
Methods of zooming an image (inclusive of converting pixel density) include selective processing conversion (SPC), in which each pixel of an original image is periodically eliminated in dependence upon the zoom ratio (this is image reduction) or in which each pixel is repeated in dependence upon the zoom ratio (this is image enlargement) (see Matsumoto, Kobayashi: “Study on the Effect of the Resolution Conversion on Picture Quality”,
The Journal of the Institute of Image Electronics Engineers of Japan
, Vol. 12, No. 5, pp. 354-362, 1983), and a method of obtaining the value of a pixel of interest by interpolation from neighboring pixels following a coordinate transformation (an affine transformation, etc.).
Further, a method proposed by the present applicant is available as a method of zooming a binary image. This method relates to an apparatus for extracting, smoothing and zooming outline information of a binary image and involves extracting outline vectors from a binary image, creating outline vectors that have been smoothly zoomed to a desired magnification (arbitrary) in the state of the extracted outline vector expression, and reproducing the binary image from the smoothing zoomed outline vectors, thereby obtaining a high-quality digital image zoomed to the desired magnification (arbitrary).
[Reduction of Data Quantity]
An example of an available method of reducing the quantity of image data when the image is stored or communicated is a coding method using a discrete cosine transform (DCT) according to the Joint Photographic Experts Group (JPEG). In another proposed method, a color image is converted to a uniform color space, after which the image is split up using color difference. The shape of a segment is subjected to chain coding, hue is coarsely quantized segment by segment using stored colors, saturation is subjected to a polynomial approximation and lightness is subjected to adaptive DCT coding, whereby the image data is coded. (See
The Transactions of the Institute of Electronics, Information and Communication Engineers
, B-I, Vol. J75-B-I, No. 6, pp. 422-430, 1992, June, 1992.)
Further, an example of a method of reducing the quantity of data in a binary image is MR/MMR coding or coding in accordance with the Joint Bi-Level Image Group (JBIG).
[Image Segmentation (Inclusive of Structuring of a Document Image)]
An example of a method of segmenting a color image is a recursive threshold-value method (see “Structuring of Color Image Information by Area Partition Processing”,
Journal of Information Processing Society of Japan
, Vol. 19, No. 12, pp. 1130-1136, Dec. 1987). According to this method, processing for deciding a threshold value for area partitioning from the value of a histogram in RGB, HSI, YIQ color features and then extracting an area is recursively repeated with regard to each extracted area. Another reported method is the limited-color expression method. This utilizes a color quantizing technique in which a color distribution is partitioned repeatedly until attainment of a number of colors in which a color error cannot be visually recognized by the human eye (see
The Institute of Electronics, Information and Communication Engineers
, 1990, Spring, All-Japan Convention, D146 7-168).
Examples of a method of segmenting image segments applied to a document image is one in which areas are judged pixel by pixel using the information possessed by neighboring pixels (see Tetsuya, Ochi: “Bilevel Rendition Method for Documents Including Gray-Scale and Bilevel Image”,
The Transactions of the Institute of Electronic and Communication Engineers
, Vol. J67-B, No. 7, pp. 781-788, 1984), and one in which an image segment is segmented by utilizing the background of a document image based upon the property that an image segment is partitioned by the background, after which the area is subjected to judgment (see Ito, et al: “Parallel Field Segmentation of Free-Format Documents”, 1979
Information Processing Society of Japan
, 20th All-Japan Convention, pp. 453-454). Some of these papers also propose applying the results obtained by image segmentation to the structuring of document images. (For example, see Yamata, et. al: “Multimedia Document Structuring Processing System”,
The Journal of the Institute of Image Electronics Engineers of Japan
, Vol. 19, No. 5, pp. 286-295, 1990).
However, the problems described below arise in the techniques described above.
[Problems with Zoom Processing]
The SPC zoom method is capable of forming a zoomed image in simple fashion but problems such as jaggies arise in terms of image quality.
Further, a zoom method using coordinate transformation is effective for natural images, such as images of scenery. However, when this method is applied to document images, the edges of character or line drawing portions such as characters and graphs become blurred owing to interpolation processing.
Further, the method of zooming a binary image proposed by the present applicant is very effective as a method of extracting, smoothing and zooming the outline information of a binary image. However, this method cannot be applied directly to zooming of a color image or grayscale image.
[Problem with Data-quantity Reduction]
The method of reducing the quantity of data by coding using a DCT is effective for a natural image. However, when a document image is compressed and reproduced, image quality markedly deteriorates at character or line-drawing portions such as characters and graphs.
The method of segmentation and coding a segment is also for application to natural images and lightness is therefore subjected to adaptive DCT coding. As a result, the lightness information comes to occupy a major part of the total quantity of information. Accordingly, this method is useful with regard to continuous tone portions of a color image. However, coding efficiency is poor with regard to documents, in which black-and-white portions or limited-color portions occupy the major part of the document and the importance of lightness information is low.
The coding method based upon MR/MMR or JBIG is effective when reducing the amount of d
Ishida Yoshihiro
Koga Shinichiro
Yoshizaki Osamu
LandOfFree
Image processing apparatus and method 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 processing apparatus and method, we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Image processing apparatus and method will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-3097846