JPH03250385A - Character string extracting system - 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 (Industrial Application Field) This invention relates to a document recognition system that reads a document using an image scanner and converts it into a coded document in order to input document information into a computer. The present invention relates to a document recognition system that outputs character strings of document items.

(Prior art and its problems) Conventionally, optical document reading devices that combine an image scanner and an image processing device have been known for inputting document information into a computer. , a method is used in which circumscribed rectangles of connected pixels are merged using a certain algorithm. In this method, if a ruled line exists, it is necessary to remove pixels forming the ruled line by preprocessing, which requires extra calculation time.

(Means for Solving the Problems) The present invention has been made in view of the above problems, and uses horizontal and vertical run lengths as reference values in a character string extraction process that extracts a character string area from a document image. This paper proposes a character string extraction method that separates and extracts character string regions from ruled lines through comparison.

(Operation) The character string extraction method of the present invention proposes a method for processing basic rectangle extraction and ruled line removal simultaneously and at high speed.

(Example) The basic idea of the present invention is that when performing OR processing of two pixel columns twice, from top to bottom and from bottom to top, the horizontal run length and vertical run length are When the run length is checked and the run length exceeds a predetermined reference value, the OR processing is not performed on the pixels forming the run. In other words, the reference value for the run length corresponds to the maximum character size, and connected pixels exceeding this value are not recognized as characters, that is, as ruled lines.

The present invention will be explained with reference to the accompanying drawings. FIG. 1 shows a general flow, FIG. 2 shows data of each step, and FIG. 3 shows a basic rectangular drawer processing flow.

In step 1 of the general flow shown in FIG. 1, an image of a document is read by an image scanner, and an image shown by a in FIG. 2 is input as data.

In step 2, OR processing in four directions is performed as basic rectangle extraction, but the order of the directions (downward, upward, rightward, leftward) can be set arbitrarily.

In this example, the order is downward, upward, rightward, and leftward (see Figure 3).6 Note that each OR process is based on the same algorithm except for the direction, so an example of the downward process is shown in Figure 4. show. FIG. 4 shows the original image before OR processing. O represents an ON pixel (indicating that there is data), and × represents a processing focus point. Here, processing is performed downward, and at the point of interest x, this pixel is not ON, but the previous pixel in the processing direction (indicated by 0) is ON, so the point of interest X becomes a candidate point. Whether the pixel at this candidate point is turned ON or not is determined according to the following two criteria.

(1) The run length (number of consecutive pixels) up to the previous pixel in the processing direction is less than the reference value.

(2) At least one of the pixels adjacent to the candidate point in the direction perpendicular to the processing direction is ON.

Here, the reference value of the run length corresponds to the maximum character size (in units of number of pixels) and is determined in advance. In the example of FIG. 4, if this reference value is set to 10, the run length at the point of interest x is 3, which satisfies the criterion (1). Furthermore, since the pixel on the left is ON, the criterion (2) is also satisfied. Therefore, the point of interest x is turned on (indicated by O, see FIG. 5). In the state shown in FIG. 5, the next point of interest X does not satisfy the criterion (2), so it is not turned on.

Next, the effect of the above-mentioned criterion (1) on the removal of ruled lines will be explained with reference to FIG. The run length up to one point before the point of interest x is 10, which is greater than (but not less than) a predetermined reference value, so it is not turned on even if the criterion (2) is satisfied. This prevents the ruled line area from being expanded. Also, the reason why it is set to be less than the reference value based on the criterion (1) is that when processing in the processing direction shown in FIG. 7 after the processing shown in FIG. This is to provide some margin for this.

The extracted basic rectangle is shown in FIG. 2b.

In step 3, a grapheme is constructed as a rectangle that circumscribes a collection of overlapping basic rectangles by projecting the basic rectangle in a direction perpendicular to the longitudinal direction of the character string (see C in FIG. 2).

In step 4, the graphemes are projected in the longitudinal direction of the character string, and a circumscribed rectangle of a collection of overlapping characters is extracted as a character string area (see d in FIG. 2).

(Effects of the Invention) Since the present invention has the above configuration, the basic rectangle extraction process consists of a simple OR process and a comparison with the run length reference value, so it can be processed at high speed. Moreover, since the ruled lines can be removed at the same time, there is no need for pre-processing for removing the ruled lines as in the conventional method, and the processing can be made more efficient.

[Brief explanation of drawings]

The drawings show one embodiment of the present invention, and Fig. 1 shows the general flow of the invention, Fig. 2 shows the data of each step, and Fig. 3 shows the general flow of the invention.
The figure is a basic rectangle extraction process flow, and FIGS. 4 to 7 are diagrams explaining the process.

Claims (1)

[Claims] A character string characterized in that in a character string extraction process for extracting a character string region from a document image, the character string region is separated from the ruled lines and extracted by comparing the horizontal and vertical run lengths with reference values. Extraction method