JPH0241566A - Electronic picture filing device - Google Patents

Abstract

PURPOSE:To facilitate retrieval by extracting areas of tables, graphics, photographs, or the like which easily remain in the human memory on a document and detecting and classifying features of their sizes, their positions, the number of areas, etc., and automatically registering them as retrieval keywords. CONSTITUTION:A document picture is converted to an electric signal by an image scanner 1, and this picture signal is written in an image memory 2. When a new document is registered in a filling device, areas having certain collective sizes of tables, graphics, photographs, or the like which easily remain in the human memory on the document are extracted, and features of their sizes, their positions, the number of areas, etc., are detected and classified and are automatically registered as retrieval keywords together with keywords which a user designates. Thus, a keyword of automatic generation can be designated based on fuzzy memorization related to tables, graphics. Photographs, or the like though keywords designated by the user are not accurately memorized at the time of retrieval of the document, and the flexibility of retrieval is increased to facilitate retrieval.

Description

DETAILED DESCRIPTION OF THE INVENTION (Field of Industrial Application) The present invention relates to automatic assignment of search keys to electronic image files.

(Prior art) When a large number of sentences and calligraphy are registered in a conventional image electronic file, the user may remember the general content and layout of the document, but may not remember the document name or keywords accurately, and may not remember the document name or keywords accurately when searching. There was a problem in that it was not possible to search for the required document because the document name or keyword could not be entered.
As a conventional technique to deal with this problem, Japanese Patent Laid-Open No. 62
In Publication No. 7415θ, reduced images are input at the same time, and the user is asked to select an image from among the reduced images during a search.

Further, in Japanese Patent Application Laid-Open No. 32-89387, an attribute image is created from the original image and the user is allowed to select or input the attribute image to calculate the degree of similarity and search.

(Problem to be Solved by the Invention) However, since the above-mentioned reduced image method (Japanese Unexamined Patent Application Publication No. 82-74156) stores reduced images, it requires a large amount of storage capacity, and it is also difficult to display reduced images. The problem remains that it takes time.

Also, the attribute image method (Japanese Patent Application Laid-open No. 82-69367)
However, there remains the problem that it takes time to input attribute images and calculate similarity.

The present invention solves these problems and allows users to
To provide an electronic image file device capable of automatically creating and assigning keywords at the time of inputting a file so that a document can be easily searched even when the keywords are not stored correctly.

(Means for Solving the Problem) The present invention scans and inputs a document, registers it in an external storage device,
In a file device for documents searched by a predetermined keyword, a means for identifying areas having a certain uniform size such as tables, figures, photographs, etc. in a document and their types, and the number and position of the identified areas. , means for detecting attributes that are easily remembered by humans, such as size, and when registering documents,
A means for registering the type and attribute of the identified area as a search keyword for the document is developed.

(Function) When a new document is scanned and input into a document image file device, the present invention scans an area of a certain group size such as a table, figure, or photograph in the document based on the type of the area. (i.e. types of tables, figures, photographs, etc.). next,
Attributes that are easily remembered by humans, such as the number, position, and size of identified areas, are detected. Then, the detected attribute is automatically assigned as a search keyword for the input document.

When registering a new document in a file device, the present invention provides a table, which is likely to remain as a pattern in human memory in the document, in addition to registering keywords specified by the user.
It extracts areas with a certain size, such as figures and photos, detects and categorizes their characteristics such as size, position, and number, and automatically registers them as search keywords. When searching for documents, even if you don't remember the exact keywords specified by the user, you can search for tables, figures, etc.
It is possible to specify the automatically generated keywords based on vague memories about photographs, etc., increasing search flexibility and making the search easier.

(Example) Figure 1 shows how, in an electronic image file device, keywords are created by extracting tables, figures, areas surrounded by frames, etc. as areas with a certain uniform size in a document. 1 shows a schematic configuration of a first embodiment of the present invention.

The document image is converted into an electrical signal by the image scanner 1 and output as binary digital data. This image signal is written into the image memory 2. Line segment memory 3 stores line figures in a document extracted from image memory 2 by CPU 4. The CPU 4 also controls the operation of the entire file device. A monitor display 6 is used to monitor input or search images. The keyboard 7 is used to input search keywords and the like.

The optical disc 8 is a large-capacity external storage device that stores input documents, keywords, and the like. The printer 9 prints out search results and the like.

In the above configuration, the operation when registering a new document will be explained first. The document is scanned and binarized and written to the image memory 2. Here, only line segment images are extracted from the document image on the image memory 2 and transferred to the line segment memory 3. For information on how to extract only line segments from an image,
A method based on the size of a circumscribed rectangular frame of connected black pixels, a method that measures the length by contour tracing, a method that uses the difference in line density,
Many methods have been proposed, such as a method using Hough transformation, and any of these methods may be used, but a simple method using run length will be explained here.

First, by scanning the image memory 2 horizontally and vertically by the CPU 4, a series of black pixels having a run length of a predetermined length or more is extracted and copied to the line segment memory 3. The correspondence between the original image 10 and the images on the line segment memory 3 is shown in FIG. This method detects line segments in the horizontal and vertical directions, so the results shown in the figure are obtained.

Next, it is determined whether each connected figure of each black pixel obtained here is an outer frame of a document, a frame of a table, or a part of another figure. FIG. 3 shows an example of a procedure for determining connected figures. This example assumes that the table has at least two rows and two columns, and also has an outer frame.

In determining a connected figure, as shown in FIG. 3, it is first checked whether the connected figure has a rectangle, and if it does not have a rectangle, the connected figure is determined to be a figure other than a table.

When the connected figure has a rectangle, check the size of the smallest rectangle, and if it is larger than the size of the smallest frame that makes up the table W H (W is the width of the rectangle, H is the height of the rectangle) When it is small, the connected shape containing the rectangle is not a table, so
A connected figure is determined to be a figure other than a table. If the size of the smallest rectangle is the size of the smallest frame composing the table W H (W
is the width of the rectangle and H is the height of the rectangle), it is checked whether the number of rectangles is one. If the number of rectangles is one, check whether the area of the rectangle divided by the area of the document is larger than a certain value A (it should be set to about 0°9), and if it is, then connect them. The figure is determined to be an outer frame. If it is smaller, the connected figure and its interior are determined to be figures other than tables. If the number of rectangles is 2 or more, there is a high possibility that it is a table, but to confirm this, check whether there are 3 or more horizontal lines whose starting and ending points have approximately the same X coordinates in the connected figure, and Check whether there are three or more vertical lines in the connected figure whose y coordinates of the starting point and end point are almost the same, and if both are YES, the connected figure is determined to be a table frame, otherwise it is a figure other than a table. It is determined that

By changing the criteria and procedures for judgment, it becomes possible to extract tables without an outer frame. As a result, each connected figure on the line segment memory 3 is labeled as shown in FIG.

Next, an example of a method for specifying the characteristics of a table will be shown. The number of independent tables is the number of tables, the area surrounded by the outer frame of the table is the size of the table, and the center coordinates of the table are the position of the table, as shown in FIG. When registering the information obtained above, such as the presence or absence of a document frame, the number of tables, the size of tables, the position of tables, the presence or absence of figures other than tables, search keys can be used in the same way as user-specified items and keywords. Register as.

When searching, for example, as shown in the example of search format settings in FIG. 6, the search can be performed using registered items set by the user and items automatically assigned by the system. Each item is a logical product (
AND) condition. Also, within each item, OR
) is also possible. Keywords and the like can also be selected by displaying a menu without the user having to input them directly from the keyboard.

It is considered that nine classifications as shown in FIG. 7 are sufficient for the position of the table, so any one can be selected from the menu. Further, since it is considered that the size is approximately the same, three levels of large, medium and small as shown in FIG. 8 may be used. In this case, small is less than 1/4 of the document, medium is between 1/4 and 1/2,
Large is 1/2 or more.

In this example, we have explained the case of searching using the set keys, but in actual use, human senses may be ambiguous, so the position, thickness, and number of tables may be changed before or after the user's set values. You may also set a search mode that allows you to expand your search. In addition, although the number of rows and columns of the table was not analyzed in this example, it can be determined by examining the number and relationship of parallel lines, so using this as a search key, the number of rows and columns of the table It is also possible to use In addition, run length detection and labeling are also performed by the CPU 4 in the explanation, but if higher speed is required, they can also be configured by a dedicated circuit.

As detailed above, in the first embodiment, the presence or absence of an outer frame of a document, the number of tables, and
Size, position, presence or absence of figures other than tables, etc. are registered in the same way as user keywords when registering images, and can be searched as one of the search keys when searching, making it easier for users to remember document name keywords correctly. Even if the document image is not searched, it can be easily searched using the above-mentioned keywords indicating features of a document image, such as a document's outer frame, table, figure, etc., as a pattern that is easy to remember, and search flexibility can be increased.

In this first embodiment, since images can be classified by attributes, it is possible to automatically edit and create data files containing only tables and drawing files containing only drawings. Furthermore, it is possible to reduce the amount of data by deleting essentially unnecessary data such as the outer frame of the document.

FIG. 9 shows a second embodiment of the present invention in which a keyword is created by extracting a halftone area such as a photographic image as an area with a certain uniform size in a document in an electronic image file device. This figure shows the general configuration of .

The document image is converted into an electrical signal by the image scanner 1 and output as a multivalued digital scanner. This image signal is processed by a character area/halftone area identification circuit (hereinafter referred to as T
/■Discrimination circuit) 20 distinguishes between a character area and a halftone area such as a photograph according to the characteristics of the image.
The halftone area is subjected to pseudo halftone processing such as dithering, and then written to the image memory 21. The mXn image (m=n is also possible) of the original image of the halftone area map 22 is mapped to correspond to one pixel, and the majority of the corresponding pixel group is identified by the T/I identification circuit 20 as the halftone area. It is set when The monitor display 6 is for monitoring input or search images. A keyboard 7 is provided for inputting search keywords and the like for storing file documents. The optical disk 8 is a large capacity external storage device. The printer 9 is used to print out search results and the like. The CPU 4 controls these operations.

The operation of the second embodiment configured as above will be explained. First, the operation when registering a new document will be explained. When a document is scanned and input, the image is binarized or dithered pixel by pixel depending on the characteristics of the image and written into the image memory 21. At the same time, writing is performed on the halftone area map 22 for pixels on the map corresponding to areas identified as halftone images of the original image 23. A halftone region is determined when the number of halftone pixels in that region (mXn pixels of the original image) exceeds the majority. It is assumed that the number of halftone pixels exceeds the majority. As shown in Japanese Patent Application Laid-Open No. 58-205376, there are many methods known as T/work identification methods, such as those that utilize differences in the density gradient of images and those that utilize differences in spatial frequency components of images. Therefore, the explanation is omitted here. FIG. 10 shows the correspondence between the original image and the image on the halftone area map. Since it is determined whether it is a halftone area or not based on the number of pixels identified as halftones in the mXn pixels of the original image, the parts corresponding to the periphery of the halftone area of the original image may be jagged on the map. There is sex. However, even if an identification error of several pixels occurs in the original image, it will not be incorrectly determined as a region.

Next, an example of a method for defining the characteristics of the halftone area will be described. For the image on the halftone area map obtained in FIG. 10, a circumscribing rectangle 24 of connected pixels is taken as shown in FIG. This area is defined as the size of the halftone area. Also, the circumscribed rectangle 2
Let the center coordinates of 4 be the position of the intermediate area. Furthermore, if there is no halftone area or if the size of each halftone area is less than a certain size (for example, it is considered that there is no halftone area smaller than 1cm x 1cm in the original image), set the number of halftone areas to 0. do. The number of areas larger than this is defined as the number of halftone areas. As described above, the obtained items, ie, the size, position, number, etc. of the halftone area, are registered as search keys in the same way as the user's specified items and keywords at the time of registration.

When searching, for example, as shown in FIG. 12, the search can be performed using registered items set by the user and items automatically assigned by the system. Each item becomes a logical product (AND) condition. It is also possible to provide a logical sum (OR) condition within each item.

It is also possible to display a menu for the user to select keywords, etc., without having to input them directly from the keyboard.

Similar to the first embodiment, it is considered that nine classifications as shown in FIG. 7 are sufficient for the position of the halftone area, so any one can be selected from the menu. In addition, since it is considered that the size is approximately the same, as in the first embodiment, the eighth
It may be in three stages of large, medium and small as shown in the figure.

In this second embodiment, a case has been described in which a search is performed using the set keywords, but since human senses are sometimes ambiguous, in actual use, the position, size, and number of halftone areas are determined by the user's setting values. You may also set a search mode that allows you to expand your search forward and backward.

As described above, in the second embodiment, the number, size, position, etc. of halftone areas obtained by the T/I identification circuit 20 are automatically detected and classified, and the user's keywords are By registering in the same way and searching as one of the search keywords when searching, even if the user does not remember the document name or keyword correctly, it will be possible to search for documents such as photos or other documents that the user vaguely remembers. It is possible to easily search using the above keywords that indicate the characteristics of the area, increasing the flexibility of the search.

Conventionally, a T/I identification circuit was used only to improve the reproducibility of halftone images, but in the second embodiment, this T/I identification circuit is used only to improve the reproducibility of halftone images.
/, Since the I identification circuit is shared, new functions can be added without considering increased costs. Also, this second
In the example, it is possible to extract the approximate halftone image area, so
When performing image editing such as scaling or image compression, an algorithm suitable for pseudo halftones can be applied, making it possible to obtain high image quality and high compression ratio.

(Effects of the Invention) When registering a new document in a file device, the present invention extracts areas of the document such as tables, figures, and photographs that are easily remembered by humans in addition to registering keywords specified by the user. However, since the features such as size, position, number, etc. are detected and classified and automatically registered as search keywords, when searching for the document, if the user does not remember the specified keyword accurately, In addition, the user can specify the automatically generated keywords based on vague memories regarding areas such as tables, figures, photographs, etc., increasing the flexibility of the search and making the search easier.

In addition, the present invention does not significantly increase the storage capacity unlike the conventional reduced image method even if the keywords are added as described above, and it does not take much time to input unlike the conventional attribute image method. I can't say that it takes.

[Brief explanation of the drawing]

FIG. 1 is a block diagram showing the configuration of a first embodiment of the present invention. FIG. 2 is a diagram showing the correspondence between the original image and the image on the line segment memory. FIG. 3 is a diagram showing an example of a method for determining connected figures. FIG. 4 is a diagram showing labeling of each connected figure. FIG. 5 is a diagram showing an example of defining the characteristics of a table. FIG. 6 is a diagram showing an example of setting a search expression. FIG. 7 is a diagram showing the relationship between positions and keywords when the positions of tables, figures, photographs, etc. are classified into nine categories. FIG. 8 is a diagram showing the relationship between sizes and keywords when classifying the sizes of tables, figures, photographs, etc. FIG. 9 is a block diagram showing the configuration of a second embodiment of the present invention. FIG. 10 is a diagram showing the correspondence between the original image and the image on the halftone area map. FIG. 11 is a diagram showing a method for defining a halftone area. FIG. 12 is a diagram showing an example of setting a search expression. 1... Image scanner, 2.21... Image memory, 3... Line segment memory, 4... CPU, 5...
Data bus, 6... Image monitor, 7... Keyboard, 8... Optical disk, 9... Printer, 20...
・Character area/halftone area identification circuit (T/I identification circuit),
22... Halftone area map. Patent Applicant Fuji Xerox Co., Ltd. Agent Patent Attorney Noboru Iwa - Figure 1 Figure 2 Figure 4 Figure 5 Figure 9 Figure 10 Procedural Amendment Form Stamp Button

Claims (1)

[Claims] In a document filing device in which a document is scanned and inputted, registered in a storage device, and searched by a predetermined keyword, tables, figures, photographs, etc. in the document have a certain uniform size. means for identifying regions and their types; means for detecting attributes that are easily remembered by humans, such as the number, position, size, etc. of the identified regions; 1. An electronic image file device comprising means for registering as a search keyword.