JPH0477965A - digital copying machine - Google Patents

Abstract

(57) [Summary] This bulletin contains application data before electronic filing, so abstract data is not recorded.

Description

DETAILED DESCRIPTION OF THE INVENTION [Field of Industrial Application] The present invention relates to a digital copying apparatus, and more particularly to a digital copying apparatus having a function of translating character information in an image from English to Japanese in units of words.

[Conventional technology] Translation work between languages generally requires very advanced knowledge.
Moreover, translation takes time. Therefore, various proposals have been made for technologies to support or automate translation work, and some technologies have even been commercialized.

For example, JP-A No. 54-154955 and JP-A No. 5
5=15562 discloses a device that translates word by word, and Japanese Patent Application Laid-Open No. 511-4480 discloses a device that performs grammar and meaning analysis to translate sentences. .

In addition, since errors still occur frequently when translating entire sentences, it is necessary to correct the translation results, and the method is not yet at a practical stage.

[Problems to be Solved by the Invention] However, in conventional translation devices, it is necessary for an operator to input character information using a keyboard or the like, for example, in the form of a word. Therefore, if there is a document in English, for example, it is necessary for the operator to read the words sequentially from the document and enter the character information manually. This type of input work requires a very large amount of effort and requires a very long time for inputting, especially in the case of an operator who is not familiar with key operations.

An object of the present invention is to simplify the input work of character information to be translated into a translation device, thereby facilitating the translation work.

[Means for Solving the Problems] In order to solve the above problems, the present invention includes an image reading means for reading a document image as a set of minute pixel information, a memory means for storing the information read by the means, and a memory means for storing the information read by the means. In a digital copying apparatus including an output means for outputting the image information stored in the memory means: a character cutting means for cutting out character information for each character from the input image information stored in the memory means: an interval between the cut out characters; On the basis of the,
Word extraction means for classifying character information into words consisting of one character or a set of multiple characters; includes a character pattern dictionary in which feature amounts for each character are registered in advance, and extracts character information from image information extracted for each character. Character recognition means for recognizing the characters indicated: includes a word dictionary in which translation results for each word are registered, and refers to the word dictionary for each word extracted by the word extraction means according to the structure of the characters constituting it. , a translation means for recognizing a translation result of each word; and an output image creation means for writing character information obtained as a translation result of the translation means into an output image area of the memory means as image information.

[Function] According to the device of the present invention, characters and words are automatically recognized from the original image, translation is executed for each recognized word, and the result is output in the form of an output image, so that the translation work is easy. Just like using a regular copy machine, it's an extremely simple process of placing the manuscript to be translated into the image scanner and pressing the copy button.

By the way, since it is very difficult to correctly recognize characters from image information, it is extremely important to increase the character recognition rate in this type of device.

Therefore, in a preferred embodiment of the present invention, a plurality of types of character pattern dictionaries are provided according to differences in character size and font, and characters to be recognized are recognized based on the line pitch and character pitch detected when cutting out characters. In addition to identifying the size of the character pattern and limiting the character pattern dictionary to be used based on the result, recognition is performed using multiple types of character pattern dictionaries for a part of the character string in the original image, and character recognition is performed based on the results. To automatically select a character pattern dictionary with a high rate and perform subsequent recognition.

When the characters in the original image are printed characters, for example, the recognition rate of the characters can be greatly improved by using a dictionary that matches the font of the characters. However, it is very difficult for the general public to recognize the font by looking at the printed text on an image, and the time required for the recognition process increases if a dictionary of multiple fonts is constantly referenced during the recognition process. . Therefore, if you refer to multiple font dictionaries for only part of the character string as described above, select only the one with the highest recognition rate for each dictionary, and use it in subsequent processing, the appropriate font dictionary will be automatically selected. can be selectively selected, the recognition rate is improved,
Moreover, characters can be recognized in a short time.

Further, in a preferred embodiment of the present invention, when the translation means cannot detect a translation result, it automatically shifts to the correction mode processing, and for each word string to be translated, a part of the characters is changed according to a predetermined rule. Make corrections based on this and run the translation again.

Since the device of the present invention automatically extracts words from one image and translates them, the extracted words may not be compatible with the structure of the word dictionary. Translation success rate decreases. For example, if there is a comma (1) in the middle of a sentence, a comma may be added to the end of the extracted word, but such words are usually not registered in dictionaries and cannot be translated. However, the generation of this kind of special word follows certain rules; for example, if the last character of a word is a comma, that character is deleted. It is possible to modify the If such automatic correction is performed, the success rate of translation can be improved without requiring correction operations by an operator.

Other objects and features of the present invention will become clear from the description of the embodiments with reference to the following five drawings.

[Embodiment] FIG. 1 shows the structure of a mechanical section of a digital copying machine of one type in which the present invention is implemented. Referring to FIG. 1, this digital copying machine is roughly divided into an upper image scanner 10.
0 and a laser printer 200 placed below it.

A contact glass on which a document is placed is placed at the top of the image scanner 100, and an optical scanning system is provided below it. A document is exposed to light by an exposure lamp 1 of an optical scanning system, and its reflected light, that is, both image lights, pass through various mirrors and lenses 2 provided in the optical scanning system and are imaged on a light receiving section 3. This light receiving section 3 is provided with a one-dimensional CCD image sensor, which will be described later. Since the optical scanning system is driven in the horizontal direction of the drawing by a mechanical drive system, the image light obtained by exposing each part of the document surface is sequentially read by the image sensor, line by line.

The image information read by the image sensor is converted into output images by the processing described later, and then sent to the laser printer 2.
The laser beam output from the writing device 4 of 00 is modulated.

The laser beam modulated by the image information passes through a writing optical system and forms an image on the surface of the photoreceptor drum 5. The entire surface of the photoreceptor drum 5 is uniformly charged to a predetermined high potential by the main charger 6 in advance, and when it is irradiated with image light, the potential changes depending on the light intensity and becomes a potential corresponding to one image. distribution, that is, an electrostatic latent image is formed.

The electrostatic latent image formed on the photosensitive drum 5 is made visible by adsorption of toner when it passes through the developing unit 7, forming a toner image.

On the other hand, recording paper is fed out from the paper feed cassette 12 or 13 selected, and the recording paper is fed to the photoreceptor drum 5.
The toner image is fed so as to overlap the surface of the photosensitive drum 5 in synchronization with the formation timing of the upper toner image. Subsequently, the toner image on the photosensitive drum 5 is transferred onto the recording paper by the urging of the transfer charger. Further, due to the biasing force of the separation charger 9, the recording paper onto which the toner image has been transferred is transferred to the photoreceptor drum 5.
The image is separated from the camera and goes to the fixing unit 14. The toner image on the recording paper is fixed onto the recording paper by the fixing unit 14, and then the recording paper is ejected from the copying machine.

After the image transfer and recording paper separation are completed, the surface of the photosensitive drum 5 is cleaned by the cleaning unit 10 in preparation for the first image formation.

FIG. 2 shows the configuration of the electrical equipment of the digital copying machine shown in FIG. 1. This will be explained with reference to FIG. image scanner 10
0, the bitmap format document image signal read by the CCD image sensor 110 is sent to the A/D
After being converted into a digital signal (8 bits in this example) by the converter 120, the shading correction unit 13
0 to correct for density level variations,
It is stored in memory unit 330. Memory unit 3
30 includes a storage area for one frame of an image, a storage area for use in character recognition and translation processing, and a storage area for output images. The translation unit 340 processes the input image information on the memory unit 330 and generates output image information on the memory unit 330 after performing character recognition and translation processing, as will be described later. A character pattern dictionary 350 is provided for character recognition, and an English-Japanese dictionary is provided for translation. The character pattern dictionary 350 includes several sets of dictionaries associated with various character types having different sizes and fonts. In the English-Japanese dictionary 360, various English words and Japanese information associated with each word are registered in pairs.

The output image information generated on the memory unit 330 is
The signal is applied to the laser printer 200 in the form of binary information corresponding to black/white of each pixel, passes through the buffer 220, passes through the LD driver 230, and is applied as an energizing signal to the laser diode 240.

Therefore, the laser diode 240 outputs laser light modulated according to the output image information. This laser light is output from the writing device 4 and is irradiated onto the surface of the photoreceptor drum 5 via an optical scanning system for writing.

Instructions from the operator are carried out by keystrokes from an operation board 310 arranged on the top of the copying machine. The main control unit 320 controls various displays on the operation board 310, reads key input from the operation board 310, and instructs each section to start reading, start hard copying one image of operation mode switching, and the like.

FIG. 3 shows an outline of the processing of the copying machine shown in FIG.

This will be explained with reference to FIG.

In step 51, the image of the document set in the image scanner is read, the image data is captured in the form of binary data indicating black/white in units of pixels, and the data is stored in the memory unit 330.
Accumulate in the upper input image area. As a result, information such as the image IMG shown in FIG. 4 is stored in the memory unit 330.
stored in bitmap format.

In step 52, in order to detect the line pitch of sentences in the input image, the distribution state of non-pixels in the vertical direction is detected in the form of a histogram HGI shown in FIG.

This histogram is used to detect the pitch of one line of text.

In step 53, character string information is extracted line by line from the input image information based on the detected line pitch.

In step 54, the distribution state of black pixels in two horizontal directions is detected in the form of a histogram HG2, as shown in FIG. 5, for each line of extracted character string information.

The pitch between characters is then detected based on this histogram.

In step 55, based on the detected character pitch,
As shown in Figure 6, from the character string information for each line, 1
Extract character information for each character.

In step 56, based on the line pitch detected in step 52 and the character pitch detected in step 54,
Detects the average size of characters included in the text to be processed.

In step 57, some dictionaries most suitable for the character size detected in step 56 are selected from among the large number of dictionaries provided in the character pattern dictionary 350. Since the character pattern dictionary 350 includes dictionaries of various fonts for each character size, dictionaries of a plurality of fonts are selected at the same time.

In step 58, the character information cut out in step 55 is processed to extract words as shown in FIG. Specifically, if the distance between adjacent character information is one character or more, the position between them is considered to be a word break, and the character information from that break to the break is considered to be the same word. Group as belonging.

In step 59, character pattern recognition is performed for each piece of character information belonging to the extracted word. That is, the feature parameters of the character information to be recognized are detected, compared with the feature parameters of various characters registered in the character pattern dictionary, and characters with similar characteristics are obtained as the recognition result of the character information. Note that in step 57, since a plurality of font dictionaries are selected, feature parameters are compared for all of them. Information indicating whether character recognition was successful or not is also stored for each font dictionary.

In step 60, an English-Japanese dictionary is referenced to find a matching English word for the extracted word consisting of a set of ig* characters. When a matching English word exists in the dictionary, the Japanese information associated with it is retrieved and a translation result is determined; otherwise, no result is determined.

If there is a translation result, the process passes through step 61 and proceeds to 64, where the result is stored in the output word storage area of the memory. If there is no translation result, it is determined in step 62 whether or not the English word to be translated can be modified. In other words, the 9th
As shown in the figure, if the English word before correction contains uppercase letters, they are converted to lowercase letters. If there is a comma at the end of an English word, remove it. S at the end of a word
If a hyphen "-" is included, delete it. If the word is a compound word connected by a hyphen, it is checked whether the word should be corrected or not based on predetermined rules, such as dividing the word into multiple parts at the hyphen. If the word can be modified, the modification is performed in step 63, and the dictionary is searched for the modified word again in step 60 to perform translation. If correction is not possible, it is regarded as a translation error and an error code is output in step 65.

In step 66, it is checked whether recognition of all characters in the first line of the sentence to be translated has been completed. If the first line is being processed, the process proceeds to step 67 because recognition has not yet been completed.

In step 67, the number of characters included in the word being processed and the number of recognized characters are calculated for each font in the character pattern dictionary used, and the results are stored.

When recognition of all characters for the first line of the sentence to be translated is completed, the process passes through steps 66 to 68 and returns to step 69.
Proceed to. In step 69, the contents of the memory holding the results of step 67 are checked, and the character recognition rate for one font is compared between fonts, and the font with the highest recognition rate is detected. Then, a dictionary for that font is selected, and from then on only this one dictionary is used for character recognition (step 59). Once the dictionary font to be used is determined, the process proceeds from step 68 to step 70, so step 69 is not processed.

In step 70, it is determined whether or not character recognition and word translation have been completed for the entire image. If not, the process returns to step 58 and the above process is repeated.

When character recognition and word translation for the entire image are completed, the process proceeds from step 70 to step 71. In step 71,
In the output image area on the memory unit 330, for example, the eighth
Output image information as shown in Figure a is created in bitmap format. That is, the memory area in which the translation results outputted in step 64 are stored is referred to, and the translation results are taken out in order, and each translation result is developed into image information representing a character pattern and sequentially written into the memory. Note that "?" shown in FIG. 8a indicates a character that could not be recognized or a word that did not exist in the English-Japanese dictionary.

In step 72, the output image information created in step 71 is output to a printer to create a hard copy of the output image.

In this embodiment, the contents of the image to be output can be switched by key operations from the operation board 310.

By this switching, it is possible to select one of the four output formats as shown in FIGS. 8a, 8b, 8c, and 8d as the horn image.

In the output image of FIG. 8b, the copy image of the input image and its translation result are arranged one above the other so that their positions correspond to each other for each line of text. The output image shown in FIG. 8c is a copy of the input image and the result of character recognition for each line of text.

and the results of its translation are arranged one above the other so that their positions correspond to each other. Similarly, in FIG. 8d, images of recognized characters and translation results are arranged line by line.

[Effects] As described above, according to the present invention, the following effects can be obtained.

Claim 1: Automatically recognizes characters and words from the original image, executes translation for each recognized word, and outputs the results in the form of an output image, so there is no need to operate the keyboard to enter characters. There is no need for translation, and the translation work is just like using a copy machine.
The process is extremely simple: place the manuscript to be translated into the image scanner and press the copy button.

Claim 2: A dictionary of size and font suitable for the characters to be recognized can be automatically selected, the recognition rate is improved, and the time required for character recognition is shortened.

Claim 3: The success rate of translation can be improved without requiring correction operations by co-operators.

[Brief explanation of the drawing]

FIG. 1 is a front view showing the structure of a mechanical section of a digital copying machine of one type that implements the present invention. FIG. 2 is a block diagram showing the configuration of the main parts of the electrical equipment section of the copying machine shown in FIG. 1. FIG. 3 is a flowchart mainly showing the contents of the processing of the translation unit 340 of FIG. 2. Figure 4 is a plan view showing the input image and a histogram of the black pixel distribution in the vertical direction, Figure 5 is a plan view showing the input image in one row and the histogram of the black pixel distribution in the horizontal direction, and Figure 6 is a plan view showing the histogram of the black pixel distribution in the horizontal direction. FIG. 7 is a plan view showing an image of characters grouped by word. FIG. Figures 8a, 8b, 8c and 8d are each 1
FIG. 3 is a plan view showing an example of a created output image. FIG. 9 is a block diagram showing the correspondence between words before correction and words after correction. 1: Exposure lamp 2: Lens 3 light receiving section 4: Writing device 5: Photosensitive drum 6: Main charger 7: Developing unit 8: Separation charger 9: Transfer charger 10: Cleaning unit 12.13: Paper feed cassette 100: Image scanner 110: CCD image sensor 120: A/D converter 200: Laser printer 310
: Operation board 330: Memory unit 340: Translation unit 350: Character pattern dictionary 360: English-Japanese dictionary

Claims (3)

[Claims] (1) A digital copying device that includes an image reading means for reading a document image as a set of minute pixel information, a memory means for storing the information read by the means, and an output means for outputting the image information stored in the memory means. In: character cutting means for cutting out character information for each character from the input image information stored in the memory means; character information for each word consisting of one character or a set of multiple characters based on the interval between the cut out characters; word extraction means for classifying; character recognition means that includes a character pattern dictionary in which feature amounts for each character are registered in advance; character recognition means for recognizing characters indicated by image information cut out for each character; Including a registered word dictionary, for each word extracted by the word extraction means,
Refer to the word dictionary according to the composition of the letters that make it up,
Translation means for recognizing the translation result of each word; and output image creation means for writing character information obtained as a translation result of the translation means into the output image area of the memory means as image information. A digital copying device. (2) The character recognition means includes a plurality of character pattern dictionaries corresponding to differences in character size and font, and determines the size of the character to be recognized based on the line pitch and character pitch detected when cutting out the character. In addition to identifying the character strings and limiting the character pattern dictionary to be used based on the results, recognition is performed using multiple types of character pattern dictionaries for a part of the character string in the original image, and the character recognition rate is 2. The digital copying apparatus according to claim 1, wherein a high character pattern dictionary is automatically selected for subsequent recognition. (3) When the translation means cannot detect a translation result, it automatically shifts to correction mode processing, corrects some characters for each word string to be translated based on predetermined rules, and then repeats the process again. 2. The digital reproduction apparatus of claim 1, wherein the digital reproduction apparatus performs translation.