JPS61190666A - High speed picture processing system - Google Patents

Abstract

PURPOSE:To save useless calculation and to enhance speed by deciding necessity and convergence of calculation making a partial picture a unit, and not making calculation on partial picture for which calculation is unnecessary. CONSTITUTION:Picture input from a picture inputting section 1 is stored in a picture memory section through a picture accumulating section 2 and a transferring section 4. An arithmetic section 6 makes decision of picture processing operation and convergence, and checking for presence of picture elements in which the value is changed. The arithmetic section 6 does not make calculation on a partial picture for which calculation is decided unnecessary. Calculated picture is outputted from the picture memory 5 through the transferring section 4, picture accumulating section 2 and a picture outputting section 3.

Description

DETAILED DESCRIPTION OF THE INVENTION [Description of the Field to which the Invention Pertains] The present invention relates to a method for controlling the range of calculations in order to process digital images at high speed.

[Description of conventional technology]

In image processing conventionally performed in the form of one-iteration local operation, convergence determination has been performed for the entire digital image. Therefore, even for parts that converged after a small number of iterations, unnecessary calculations were performed, increasing the calculation time.

Also, "Only pixels near the pixel whose value has changed.

Due to the influence propagation rule that states that the value may change next time, unnecessary operations are performed on pixels whose values do not change even if operations are performed, increasing the calculation time.

As a high-speed image processing method to solve the second problem, (
! It has been proposed that the coordinates of pixels whose values have been changed are stored, and pixels in the vicinity thereof are subjected to ninth-order calculations. However, in this method, (1) the area for storing coordinates may become large depending on the content of the digital image to be processed;

■If the entire digital image cannot be stored in the main memory, the transfer from the auxiliary memory to the main memory is inefficient, and the 0-image scanning method does not allow processing that may affect the processing results. The problem was that it was difficult to apply.

[Purpose of the invention]

In order to solve these drawbacks, the present invention divides a digital image into a set of n pixels (partial images) that appear continuously when nine images are scanned, and calculates the convergence or necessity of calculation for each partial image. Determine.

Partial images that no longer require calculation are excluded from calculation targets.

Below, a digital image sampled on an orthogonal grid is
Using TV-style image scanning that advances one row from the top row to the bottom row, the new value of the pixel in the i-th row and j column is
Row j-1, column i-1, column j, row i-1, column j+1, row i, column j-1, row i, column j, row i, column j+1, column i+1
Row j-1 column.

The process of obtaining the values from the values of the 4th+1st row, jth column and the i+1th row, j+1st column will be explained in detail with reference to the 9 drawings. In addition,
Assume that the digital images are numbered 1-M from the top row to the bottom row. Further, the two-iteration type local operation can be further classified into parallel type and sequential type. Parallel type means that in the second image scan.

The value of each pixel is calculated only from the results of the -1st image scan. On the other hand, the sequential type means that in the second image scan, the value of each pixel is calculated, a new value is calculated for the pixel for which a new value has already been calculated, and a new value is calculated for the other pixels from the result of the first image scan. It is something to do.

[Description of the structure and operation of the invention]

FIG. 2 shows two examples of the present invention. A method of dividing a digital image into partial images is shown using an example of a digital image with a size of 8 rows and 12 columns.

In FIG. 2, one partial image is a set of six pixels that appear continuously when a digital image is scanned in a TV-style manner. The smallest rectangle represents one pixel, and the number within it represents the number of the partial image to which the pixel belongs. For example, a pixel written as 1.2 indicates that it belongs to the second partial image from the left in the i-th row. in this way.

When scanning an image while advancing one line at a time, one line is divided into several parts and the resulting images are called partial images. An effect of this is that the management of one partial image becomes easier.

FIG. 3 illustrates partial images necessary when calculating the m-th partial image from the left in the i-th row. The smallest rectangle represents a partial image, and the partial image indicated by diagonal lines is the partial image required when calculating the m-th partial image from the left in the i-th row. When the m-th partial image from the left in the i-th row is at the left end, the m-1-th partial images in each of the i-1st row, the i-th row, and the i+1th row are not needed. When the m-th partial image from the left in the i-th row is at the right end, the m+1-th partial images in each of the i-1st row, the i-th row, and the i+1th row are not needed.

FIG. 4 illustrates a table of flags indicating whether or not to perform calculations on each partial image, taking FIG. 2 as an example. The row number and column number in the table of FIG. 4 correspond to the row number of the digital image in FIG. 2 and the number of the partial image from the left, respectively. On each flag on the table.

If the corresponding partial image should be subjected to an operation, "true" information is given, and in other cases, "false" information is given. Three tables are prepared: a first table, a second table, and a third table. The k-th image scanning control table A is defined as the first table. Further, the second image scanning control table B is defined as the second table if k is an odd number, and as the third table if k is an even number.

When starting the first image scan, all flags in the first image scan control table A and the first image scan control table B are set to "true". Also.

When starting the kth image scan, all flags in the +11th image scan control table B are set to "false". k
During the second image scan, process the partial image for which both the flags in the second image scan control table A and the second image scan control table B are "true", and perform the following processing on the result (1) and process (2).

(]) Check whether the values of pixels included in the partial image will be changed in subsequent calculations. About this.

It can be determined from values near each pixel according to conditions depending on the image processing function. If all the pixels have converged, the flag of the partial image in the second image scanning control table A is set to "false".

(2) Check whether there are any pixels whose values have been changed. l(+1
The flag of the image scanning control table B is set to "true" for the second time. In the case of sequential operations, when these partial images are detected, the flag in the second image scan control table B is set to "true" for partial images that have not been reached by the second image scan. .

The effect of this is that it is possible to determine the partial images that need to be subjected to calculations, thereby eliminating unnecessary calculations. Moreover, it has the characteristic that it can be realized by using only a certain amount of storage capacity and adding relatively simple processing. Note that only one of the above processing (1) and processing (2) may be executed.

If only process (1) is performed, the two-image scanning control table B is not required. When only processing (2) is performed, the two-image scanning control table A is not necessary.

FIG. 1 shows an embodiment of the present invention. l is an image input section, 2 is an image storage section, 3 is an image output section, 4 is a transfer section, 5 is an image memory section, 6 is a calculation section, and 7 is a transfer section. A control table section, 8 is an image scanning control table section, and 9 is a control section.

The image input unit 1 samples and quantizes one image.

This is the part that converts it into a digital image and inputs it.

These include a facsimile, a drum scanner, and a TV left camera. The image storage section 2 is a section that stores digital images provided from the nine image input sections 1, and is composed of a magnetic disk device, its control section, and the like.

The image output section 3 is a section that displays the digital image stored in the one-image storage section 2, and is a printer, a display, or the like.

The transfer unit 4 transfers a part of the digital image stored in the 1-image storage unit 2 to the image memory unit 5 in units of partial images, and transfers a part of the digital image stored in the 1-image memory unit 5 in units of partial images. This is the part that transfers the image to the image storage section 2.

The image memory section 5 is a section that stores a portion of the input image in each image scan and a portion of the output image that is the calculation result, and has a storage area for five lines of the digital image to be processed. The configuration of the image memory unit 5 is such that when a calculation is performed on the first row, the area for each one row contains the i-th row, i-1st row, i+1th row, and i-th row of the input image in each image scan. The i-th row and i-1th row of the output image, which is the calculation result, are placed. However, instead of storing all partial images in each row, for the input image (as described in Figure 3)
Only those necessary for the calculation, and only the output image that has been subjected to the calculation, are placed in the corresponding position. In addition, in the case of a sequential operation, the operation results of the partial images in the i-th row and the i-1th row are written to the areas that store the i-th row and the i-1th row of the input image, respectively. Therefore, the one-image memory unit 5 can be configured with a storage area for three lines of the digital image to be processed.

The arithmetic unit 6 is a part that performs two-image processing operations on one partial image in the one-image memory unit 5, determines convergence, and checks whether there is a pixel whose value has changed. The transfer control table section 7 is a section that stores the numbers of partial images placed in each row of the one-image memory section 5. The image scanning control table section 8 is a section that stores a table of flags indicating whether or not each partial image described in FIG. 4 is to be processed. Control unit 9
is the transfer unit 4. Arithmetic unit 6. Transfer control table section 7° This is a section that controls the image scanning control table section 8.

This control procedure is shown in FIG. In addition.

When the storage area of the image memory section 5 is for two digital images to be processed, the storage area of the two image storage section 2 and the image memory section 5 can be divided by alternately using the storage area for the input image and the output image of each image scan. Transfer control can be made into an hour wheel.

FIG. 5 shows the control procedure of the control section 9 shown in FIG.

This is shown for one image scan. By repeating this procedure a given number of times or until there are no more partial images to be computed in one image scan, five iterative local computations can be performed. below.

The process will be explained according to the processes P1 to P15 shown in FIG.

(Pl) Initialize line number i to 1.

(P2) Initialize the instruction number m of the partial image to 1.

(P3) The one-image scanning control table unit 8 checks whether the m-th partial image in the i-th row should be subjected to calculation.

(P4) If it is determined that no calculation is necessary.

Control is transferred to the processing PIO.

(P5) The transfer control table section 7 checks whether the partial image of interest and the partial images necessary for its calculation are placed in the one-image memory section 5, and if the partial image does not exist, one image is stored. From storage section 2 to image memory section 5
The transfer unit 4 is instructed to transfer the information to the transfer unit 4.

(P6) When the transfer unit 4 notifies the end of transfer, the number of the transferred partial image is written in the transfer control table unit 7.

(Pl) Give the number of the partial image to the calculation unit 6 and instruct the calculation.

(P8) Receives the determination result of the convergence of the partial image and the inspection result as to whether or not the value has changed from the calculation unit 6, and according to the rules described in the explanation of FIG. Create/modify 8.

(P9) The number of the partial image is written into the transfer control table section 7 as the partial image subjected to the calculation.

(PIO) Check whether the (m+1)th partial image of the i-th row exists. If it exists, go to process P11,
If it does not exist, control is transferred to process P12.

(pH) Add 1 to the instruction number m of the partial image, and process P3
Return control to.

(P12) Obtain the number of the partial image on which the calculation was performed in the t-Lth row from the transfer control table unit 7.

These partial images are transferred from one image memory unit 5 to the image storage unit 2.
The transfer unit 4 is instructed to transfer the information to the transfer unit 4.

(Pl3) Check whether the i+1th row exists.

If it exists, control is transferred to process P14, and if it does not exist, control is transferred to process P15.

(P14) Add 1 to the row number i and return control to process P2.

(P15) Obtain the numbers of the partial images subjected to the calculation in the first row from the transfer control table section 7, and transfer those partial images from the image memory section 5 to the image storage section 2. Instruct the transfer unit 4.

[Explanation of effects]

As explained above, according to the present invention, the necessity or convergence of the calculation is determined in units of two partial images in the image processing method using the two-repetition local calculation method.

Since calculations are not performed on partial images that no longer require calculations, unnecessary calculations can be omitted.

This has the advantage of shortening calculation time. for example.

Size 1000 lines x 100 obtained from 2 basic land maps using this high-speed image processing method (n = 10 to 20)
The calculation time for thinning the 0th column digital image using the so-called H41ditch method is 48%, and the calculation time for line shape distance conversion for a 120 row x 120 column digital image obtained from a chest X-ray image is 59%. be shortened. Also, the one-piece method.

The present invention can also be applied to Hilditch's line thinning method, which can be implemented using a memory with a certain storage capacity, and in which the image scanning method affects the processing results. Furthermore, as shown in the embodiment, when the entire digital image cannot be stored in the main memory, it can be efficiently transferred from the auxiliary memory to the main memory.

[Brief explanation of drawings]

Fig. 1 is a block diagram of an embodiment of the present invention, Fig. 2 is a method of dividing a digital image into partial images, Fig. 3 is a partial image necessary to perform an operation on one partial image, and Fig. 4 is an image of an image. Scanning Control Table FIG. 5 is a diagram showing the control procedure of the control section in the embodiment of FIG. 1. DESCRIPTION OF SYMBOLS 1... Image input section, 2... Image storage section, 3... Image output section, 4... Transfer section, 5... Image memory section, 6
...Arithmetic section, 7...Transfer control table section, 8...
Image scanning control table section, 9...control section. Patent Applicant: Nippon Telegraph and Telephone Public Corporation Patent Attorney Mori 1) Kanshichip#＾ITokangetsusen＄2 Possible

Claims (1)

[Claims] For each pixel, which is a constituent unit of a digital image, the process of calculating a new pixel value from the value of the pixel in the vicinity of the pixel of interest and including the pixel of interest is performed until a predetermined condition is met. In an image processing method that is performed repeatedly, a digital image is divided into a set of multiple pixels that appear consecutively when all pixels are scanned one by one, and these are treated as partial images and calculations are performed in units of partial images. A high-speed image forming apparatus characterized by comprising means for determining the convergence or necessity of each calculation each time a calculation is performed, and means for excluding partial images that no longer require calculation from the calculation target according to the judgment result by the judgment means. Image processing method.