JPS63100873A - Method and apparatus for binarization - Google Patents

Abstract

PURPOSE:To reproduce an excellent binary picture by using a random number having plural of variance depending on the property of a picture. CONSTITUTION:An input signal Ixy being density information at a coordinate (x, y) is fed to a comparator 22 and a prediction error signal generating 2nd circuit 24b of an error detector 24. On the other hand, nearby picture signals Ix-1, y and Ix, y-1 near the coordinate (x, y) are fed to a prediction signal generat ing 1st circuit 24a to output prediction signal (Ix-1, y+Ix, y-1)/2. The 2nd circuit 24b outputs a prediction error signal exy from the input signal Ixy and the predic tion signal and uses a conversion table to convert the signal into, e.g., 6 stages of any size of selection parameter (p), the result is inputted to a regular random number converter 20b of a regular random number generator 20.

Description

DETAILED DESCRIPTION OF THE INVENTION (Field of Industrial Application) This publication EJ1 relates to a method and apparatus for binarizing a multivalued image signal representing the density of an image using random numbers.

(Conventional technology) Conventionally, this type of technology has been described in the literature 1 "Information Processing" Vol.
'15. No, 7. P, 503-509 (July 1974
(Mon.) Fujimura Kore 1J ``Method of displaying C light figures using a dot-type display device''. First, this prior art will be explained.

FIGS. 4A and 4B are block diagrams for explaining a configuration example of this conventional binarization method and apparatus. Now, assuming that the image is in the (x, y) coordinates, the input signal I
'XV is a density signal of a pixel at a certain position in the image, for example, coordinates (x, y), and is a multivalued signal that can take various values between O≦IXV≦IlaX. The random number r generated by the random number generator is a uniform random number satisfying 0≦r≦I□.

In the configuration example shown in FIG. 4(A), the random number r is used as a spy, and the human input signal IXV is compared with the random number r generated from the random number generator 10 at a comparison 'jA12, P, , = l I
This is a technique for outputting an output signal P by converting it into two by the promise F of xv≧rO, y<r.

On the other hand, in the configuration example shown in FIG. 4 CB), the input signal I×ν
And a random number is generated! After adding the random number r from 810 in the adder 14, the comparator 12 compares it with the fixed value (/iT, Pxv=l I xv+ r≧To IXV
This is a technique that binarizes and outputs an output signal pxv under the promise that +r<T.

(Problems to be Solved by the Invention) However, the conventional technology is a technology that binarizes without considering changes in image density (C light), and is suitable for reproducing halftone images; When it comes to reproducing line drawings such as lettering figures, there is a drawback in that they are reproduced as broken line segments. Furthermore, there is a problem in that an overall blurred image is obtained even for halftone images. This blurring problem can be improved by binarizing the image after edge enhancement, but parts other than the edges contain noise, making it difficult to obtain a clear image.

The purpose of this IJI is to provide a binarization method for reproducing line drawings and improving halftone blur, as described above, and an apparatus for implementing this method.

(Means for solving the problem) In order to achieve the purpose of this IJJ, according to the method of this IJJ, random numbers with small variance are used for areas with large density changes in the image. Perform value conversion processing, and
Binarization processing is performed using random numbers with large variance for image parts where density changes are small. A prediction signal is formed from the neighboring image signals, a prediction error signal representing a density change is formed from this prediction signal and the image signal representing the density at a certain position, and this prediction 2; It is preferable to configure the random number with a variance to be selected from among a plurality of types of random numbers with different variances.

Furthermore, in order to achieve the purpose of this bulletin 11,
According to the device of No. 1, in addition to the comparator, an error detector and a normal random number generator are provided. This error detector is configured to generate a prediction error signal representing a change in density from an image signal representing the density at a certain position in the image and a neighboring image signal representing the density in the vicinity thereof. Also, the normal random number generator is
The configuration is such that a random number having a fraction corresponding to this prediction error signal is selected from a plurality of types of random numbers with different variances, and this random number is output to the comparator as a normal random number. In this case, when the prediction error signal is large, a random number with a small variance is selected, and when the prediction error signal is small, a random number with a large variance is selected and output.

In carrying out this second invention, the error detector is provided with a first circuit that generates a prediction signal that predicts an image signal of the above-mentioned degree δ from neighboring image signals, and this prediction signal and
It is preferable to provide a second circuit that generates a prediction error signal from the above-mentioned image signal at a certain position.

Furthermore, in implementing this device, a uniform random number generator is provided to oscillate uniform random numbers, and the received --like random numbers are distributed according to the cumulative curve selected corresponding to the prediction error signal. It is preferable to provide a normal random number converter for converting to a certain normal random number.

(Function) As described above, according to the method and apparatus of the present invention, as a binarization method, random numbers with small variance are used for the edge portion of characters, line drawings, or halftone images. This is a branching technique that converts the values into values and unifies them using random numbers with a somewhat large variance for the intermediate tones. Therefore, it is possible to perform binarization processing of a multivalued image signal by setting rice planting according to the magnitude of the change in image density, so that a good binary image can be reproduced.

(Embodiments) Hereinafter, embodiments of the present invention will be described with reference to the drawings.

FIG. 1 is a block diagram showing an example of the configuration of a binarization device of the present invention. In the figure, 20 is a normal random number generator, 22
denotes a comparator and 24 denotes an error detector, respectively.

The error detector 24 receives as an input signal an image signal Ill representing the density at a certain position in the image in the xy coordinate system (coordinates are (x, y)). On the other hand, a nearby image signal representing a density near this point 21 (x, y) is detected and supplied to the error detector 24. In this embodiment, image signals at adjacent positions in the X and y directions are used as neighboring image signals IX, Vl and IA.
Let it be l, y.

This error detector 24 has these neighboring image signals Ix, y
A first circuit 24 that predicts the image signal at the coordinate (x, y) (7) position from l, IXl, and V and generates a predicted signal.
Provide a. In this embodiment, this first circuit 24a includes, for example, Ix 1. It is configured to output a predicted signal with a magnitude of V + IX, V 1.

Furthermore, in this embodiment, the error detector 24 includes the first circuit 2.
A prediction error signal e is obtained from the prediction signal from 4a and the input signal.
A second circuit 24b that generates XV is provided.

However, the prediction error signal eXV is, for example, shall be.

Furthermore, in this embodiment, this error signal eXV is converted into a selection parameter p divided into a plurality of stages according to its magnitude according to a conversion curve as shown in FIG. In addition, in FIG. 4, the prediction error signal eXV is plotted on the horizontal axis, and
The selection parameter p is plotted on the vertical axis. This conversion can be performed, for example, by table conversion using a selected parameter converter 24c configured using a memory, or by any other suitable method. This selection parameter p is a parameter for selecting an accumulation curve of uniform random numbers used for selecting a normal random number from uniform random numbers in a normal random number converter to be described later.

The normal random number generator 20 includes a uniform random number generator 20a that generates uniform random numbers at an arbitrary generation rate, and a regular random number generator that converts this uniform random number into a regular random number r according to an accumulation curve corresponding to the aforementioned selection parameter p. and a converter 20b. These uniform random number generator 20a and normal random number converter 20b can be easily configured using memories storing random number tables.

FIG. 2 is a diagram for explaining an example of a +E random number conversion table stored in the normal random number converter 20.

In the figure, the horizontal axis shows the normal random number r, and the vertical axis shows the uniform random number r, and each curve numbered 1 to 6 in the figure shows the variance of the .-like random number. Cumulative curves for different normal distributions, each of which is a transformation curve of uniform random numbers versus normal random numbers with a desired variance. Which cumulative curve to select from among these multiple (six in this embodiment) cumulative curves is determined by the magnitude of the selection parameter P described above.

This selection parameter p is divided into stages corresponding to the number of cumulative curves, and the cumulative curves correspond to those with smaller variances in the order of numbers 1 to 6 in the figure. The cumulative curves are divided for each parameter so that when the parameter p is large, a cumulative curve with a small variance is selected, and when the parameter p is small, a cumulative curve with a large variance is selected. Further, the broken line in the figure is a cumulative curve of uniform random numbers, and the cumulative number in this case is normalized by I@aX (However, I sax = 255).

Next, this episode I! A theory regarding the operation of Il's 1 moeification device [1
do. The input signal IXV, which is C degree information at a certain coordinate position (:<, y), is input to the comparator 22 and the error detector 24 at
The measurement error signal is supplied to the first circuit for generation 24b.

force, the nearby image signal IX1 at a nearby position of this position (x, y). V, Ix, vl are supplied to the f measurement signal generation circuit 24a, and the measurement signal lx1. Outputs V + IXI/1. The second circuit 24b outputs the above-mentioned prediction error signal eXV from the human input signal IXV and the prediction signal,
This is converted into a selection parameter p having one of six sizes, for example, using a characteristic conversion table as shown in FIG. 4, and inputted to the normal random number converter 20b of the normal random number generator 20. .

・Uniform random numbers are always generated from the random number generator 20a,
It is input to the normal random number conversion 120b. When the selection parameter p is input to the normal random number converter 20b, a corresponding cumulative curve is selected. For example, assume that the cumulative curve number 1 is selected. At this time, the uniform random number r input to the normal random number converter 20b is converted into the corresponding normal random number r according to the selected cumulative curve No. 1. This converted normal random number r is supplied to the comparator 22 as an output. In the comparator 22, comparison processing is performed using this normal random number r as a dark value, or unlike this, the input signal IX
The value obtained by adding a normal random number to V may be compared with another fixed m value. Regardless of which method is used, it is possible to output a 2-1 output signal pxv of a multiple-1 image signal. .

It will be clear that the invention is not limited only to the embodiments described, but can be subjected to many variations and modifications.

(9, Ming effect) J: From the theory 11N mentioned above, it is clear that IJI
In JIJ's method and equipment, random numbers with multiple types of variance are used depending on the nature of the image, so for objects with large 2 degree changes, such as character Φ line figures, random numbers close to fixed insets with small variance are used. It is possible to convert grayscale images into binarized images using random numbers with a large variance, and it is possible to reproduce good 1n pli images. Furthermore, this method has an effect of emphasizing areas where the C degree change is large even in a grayscale image.

[Brief explanation of the drawing]

FIG. 1 is a block diagram showing an example of a device configuration for explaining the binarization method and t-value conversion device of this IJI, and FIG. 2A shows a normal random number conversion table used to explain this IJI. Figure 3 is a FILM error signal vs. selected parameter conversion curve diagram for IJI of this invention; This is a second diagram. 20... Normal random number generator, 20a... Uniform random number generator 20b...+F random number converter, 22... Comparator 24... Error detection base 24a... (Preliminary A leg signal For generation) first circuit 24b.
...(For prediction error signal generation) Second circuit 24c...Memory. Patent Applicant: Oki Electric Industry Co., Ltd. Official 3rd Edition New Year's Drinking Book Figure 2

Claims (5)

[Claims] (1) When comparing the image signal representing the density of the image with a threshold value using random numbers and binarizing it, random numbers with small variance are used for image parts with large density changes, and the density A binarization method characterized by using random numbers with large variance for image parts with small changes. (2) Prepare multiple types of random numbers with different variances, form a predicted signal from nearby image signals from a certain position in the image, and use the predicted signal and the image signal from the certain position to The binarization method according to claim 1, characterized in that a prediction error signal representing a change is formed, and a random number of variance is selected according to the value of the prediction error signal. (3) In a binarization device that includes a comparator that binarizes an image signal representing the density of an image by comparing it with a threshold using a random number, the image signal from a certain position in the image and the vicinity of the certain position are used. an error detector that generates a prediction error signal representing a density change from a neighboring image signal from a neighboring image signal; selecting a random number with a small variance from a plurality of types of random numbers with different variances when the prediction error signal is large; A binarization device comprising: a normal random number generator that selects a random number with a large variance when the prediction error signal is small and outputs the selected random number as a normal random number to the comparator. (4) The error detector includes a first circuit that predicts the image signal at the certain position from the neighboring image signal and generates a prediction signal, and a first circuit that generates a prediction error signal from the image signal at the certain position and the prediction signal. 4. The binarization device according to claim 3, further comprising a second circuit for generating the signal. (5) The normal random number generator includes a uniform random number generator, receives uniform random numbers from the uniform random number generator, and distributes the uniform random numbers according to an accumulation curve selected in response to the prediction error signal. The binarization device according to claim 3 or 4, further comprising normal random number conversion for converting into a certain normal random number.