JPS6331826B2 - - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION The present invention relates to an information retrieval method for retrieving information stored in a memory.

In general, when reading out information stored in memory as needed, it is necessary to search for information in memory, and the search methods usually include keywording and classification according to the recorded content. I try to code it in advance. However, when using such information retrieval methods, for example, when assigning keywords, if you do not fully understand the content of the information and then assign optimal keywords, not only will it become impossible to search later, but it will also be difficult for individuals to assign keywords. This has resulted in differences and inefficiency.

As a solution to this problem, attempts have been made to automatically extract appropriate nouns that are used neither too often nor too often from text information to perform information retrieval. There are, but
The current situation is that fully satisfactory results have not been obtained.

Furthermore, a fundamentally difficult problem with conventional information retrieval methods is the retrieval of pattern information. That is, in a memory that stores image information, for example, a specific figure existing in a document image,
It has become extremely difficult to search for photos and the like using keywords.

The present invention has been made in consideration of the above points, and
In particular, the present invention provides an information retrieval method that allows specific pattern information to be easily and quickly retrieved from a memory that stores image information as white and black density level information in pixel units.

Hereinafter, one embodiment of the present invention will be described in detail with reference to the accompanying drawings.

FIG. 1 shows an example of a configuration for concretely implementing the information retrieval method according to the present invention. an image memory M in which binary quantized density level information is stored; a buffer memory BM in which the contents of the memory are sequentially read out pixel by pixel and information for a specific image area is temporarily stored collectively; a first run length distribution generator RLF1 that calculates a run length distribution function for a specific pixel region;
A pattern generator PG that generates a pattern image to be searched, and a second run length distribution generator that calculates a run length distribution function of the generated pattern.
RLF2, register RG that stores the result,
It is constituted by a comparator CMP that performs a search by comparing the comparison standard set in the register RG with the output of the first run length distribution generator RLF1. Note that normally, binary images are often stored in the image memory M in a run-length encoded and data-compressed state, but here the non-encoded binary images are stored as they are. shall be.

Now, a case will be described in which a search is performed to determine whether or not pattern information as shown in FIG. 2 exists in the image memory M.

In the present invention, first, the pattern information shown in FIG. 2 to be searched is generated by the pattern generator PG, and the run length frequency distribution is calculated by the run length distribution generator RLF2 according to the generated pattern. . At this time, using the pattern information shown in FIG. 2, a run length distribution function F(l) relating only to the run length of white pixels is obtained as shown in FIG. 3, and the distribution function F(l) is set in the register RG. Further, the density level information is sequentially read out from the image memory M pixel by pixel and stored in the buffer memory.
It is sent to BM, where it is two-dimensionally rearranged and M
Pattern information in a specific image area having a ×N pixel configuration is extracted. Next, the run length distribution function F′(l) in the extracted specific image area is calculated by the run length distribution generator RLF1,
The result F'(l) is sent to the comparator CMP where the comparison reference F(l) is preset in the register RG.
A comparison is made with At that time, the comparator CMP calculates the difference or correlation coefficient between F′(l) and F(l), and if the comparison result is within a preset threshold, the image at that time is It is determined that the pattern information in the specific image area extracted from the memory M has been retrieved, and the determination result is sent to an external display device or the like.

FIG. 4 shows the first run length distribution generator.
This shows a specific example of the circuit configuration of RLF1.
Clock from buffer memory BM with ×N pixel configuration
Pixel information read out in synchronization with CLK is sequentially sent to shift registers SR1 and SR2 for one pixel, and the output of each shift register SR1 and SR2 is sent to a comparator.
A comparison is made by CMP to determine whether or not the contents of the two match. If they match, a count command is issued to the counter CNT to count the run length. At that time, if the comparison contents by the comparator CMP do not match, the contents of the RAM storing the run length R up to that point are read out using the output of the counter CNT at that time as an address and sent to the adder ADD. , then 1 is added and it becomes the run length W of the white pixel again.
Written to the same address in RAM. In this way, the run length distribution function F'(l) of pattern information within a specific image area can be easily determined. In addition, a second run length distribution generator
The configuration of RLF2 is exactly the same.

Note that in the above embodiment, the pattern information to be searched and its reference pattern information are compared based on the frequency distribution of run lengths, but the present invention is not limited to this in any way. It goes without saying that the search may also be performed by comparing the run lengths of each scanning line in an M×N pixel area, for example. In that case, each of the first and second run length distribution generators RLF1 and RLF in FIG.
2 would be replaced by providing a run length counter for each scan line. However, in this case, as a matter of course, it cannot be denied that the pattern image becomes susceptible to sampling errors (positional errors of reading pixels). Furthermore, when the scales of the pattern information to be searched and the reference pattern are different, the sampling density (size in pixel units) may be changed in advance or as appropriate so that the scales of both approaches the same value in terms of run length. Needless to say.

As mentioned above, regarding the information retrieval method according to the present invention,
When retrieving specific pattern information from a memory that stores images as pixel-by-pixel binary information while comparing it with reference pattern information, run length is now used as a parameter to compare and determine whether the two match. It has the excellent advantage of making it possible to easily and quickly search for pattern information such as figures and photographs, which has been difficult in the past.

[Brief explanation of the drawing]

Fig. 1 is a block configuration diagram for concretely implementing the information retrieval method according to the present invention, Fig. 2 is a diagram showing an example of pattern information to be searched, and Fig. 3 is a characteristic diagram showing its run length frequency distribution. , FIG. 4 is a block diagram showing a specific example of the configuration of the run length distribution generator in the embodiment. M...Image memory, BM...Buffer memory, RLF
1, RLF2...Run length distribution generator, PG...Pattern generator, RG...Register, CMP...Comparator,
SR1, SR2...Shift register, CLK...Clock, CNT...Counter.

Claims (1)

[Claims] 1 When searching for specific pattern information while comparing it with reference pattern information from a memory that stores images as binary information in pixel units, run length is used as a parameter to compare and determine whether the two match. An information retrieval method characterized by the following.