JPS58166874A - Picture processing system - Google Patents

Abstract

PURPOSE:To obtain an easy-to-see recorded picture, by reading the picture with small and large scanning points, multiplying the signal subtraction value by the contour emphasis coefficient which is varied in response to the picture density and adding the signal obtained at a small scanning point. CONSTITUTION:The image of an original undergoes a photoelectric conversion through a slit having a normal level of image resolution and another slit larger than the former. A subtraction is carried out by an operational amplifier 10 between the signal which is read in a normal size and passed through an LPF 9 with elimination of the noise of the high frequency component and the blur signal which passed through an LPF 8 the noise of the high frequency component. These subtracted signals are fed to a multiplier 13. While a function generator 12 feeds the outputs of LPFs 8 and 9 and applies the output to the multiplier 13. The output signal given from the multiplier 13 is supplied to an adder 11 and added with the signal passed through the LPF 9 to be delivered in the form of a contour emphasizing signal. In such a way, both light and strong tones of picture can be partially expressed to obtain an easy-to-see recorded picture.

Description

DETAILED DESCRIPTION OF THE INVENTION The present invention reads an image with large scanning points and small scanning points;
The present invention relates to an image processing method that performs edge enhancement etc. using these signals.

Image contour enhancement is an image processing method for obtaining sharp images from originally blurred image data. Further, the image data contains unnecessary information (noise), and a method for removing this information is noise removal of the #J image. Emphasizing the contours of an image means further emphasizing the parts of the image where the two-dimensional density changes are significant, and from a different perspective, emphasizing the high frequency components of the image. It is. On the other hand, noise in image data generally includes high frequency components, and noise removal, which is commonly known, also means removing these high frequency components. In other words, since the purpose is to impart blur, these two image processes (contour enhancement and noise removal) are contradictory, and it is difficult to achieve both with a simple processing method.

As an example, consider a system that scans to read and record images, such as a facsimile machine.

Image data noise in such devices can be broadly classified into two types. One is noise generated in the photoelectric conversion unit that reads the image (fluctuations in the light source and noise from the photoelectric conversion element), and the other is noise inherent in the document itself. The former is noise generated during the process of scanning the manuscript to obtain image data, but the latter is information outside of the information that the observer needs from the manuscript.
Possible causes include dust or scratches on the original, or roughness due to the paper quality of the original. Regarding the nature of the noise, the former occurs in one dimension, and the latter exists in two dimensions. Among these types of noise, noise of -order low frequency components and noise having a two-dimensionally wide area cannot be removed without performing complex processing such as pattern identification.・
・ Concerning high-frequency components and isolated noise in a narrow area, the conventional contour enhancement processing that is commonly performed predicts the frequency components of the noise to a certain extent and removes them or adds a slight blur. Contour enhancement is often performed afterwards. In short, by knowing where the frequency components you want to extract as contours are distributed and applying a pass filter for that band, you can detect and enhance the contours of the image.

An example of conventional contour enhancement will be explained below based on the drawings.

FIG. 1 is a block diagram of a conventional edge enhancement device, and FIGS. 2a to 2e show output waveform diagrams of each block in FIG. 1. The image of original 1 is captured by lens 2 and half mirror 3.
The image is formed on the slit 4 and slit 60 planes through the beam. The slit 4 has a size that allows the image of the original 1 to be divided into regular resolutions, and the slit 6 is made larger than the slit 4. The size ratio is determined □° in consideration of the method of contour enhancement, and is determined to the extent that the image after contour enhancement does not look visually unnatural. The slit 6 on the larger hole side is also called a blur mask. Naturally, the optical image of the hole in the slit 6 and the optical image of the hole in the slit 4 are perfectly matched in their center positions on the image of the document 1, and the surface of the document 1 is scanned like a facsimile machine. I will continue to do so. It is considered that the inside of the double circle on the surface of the original 1 corresponds to the slit 4 hole, and the outside of the double circle corresponds to the slit 6 hole. The light passing through the slit 4 enters the photoelectric conversion element 5, and the light passing through the slit 6 enters the photoelectric conversion element 7, where it is converted into an electrical signal. The output waveform of the photoelectric conversion element 6 is shown in Figure 2 (
In a), the output waveform of the photoelectric conversion element 7 is shown in FIG. 2(b). The amplitudes of these two signals are optically
or shall be electrically regulated. The output signal of the photoelectric conversion unit +5 passes through a low-pass filter 9, and the output signal of the charging conversion element 7 passes through a low-pass filter 8, resulting in a signal from which high-frequency component noise has been removed. A waveform obtained by superimposing each signal is shown in FIG. 2(c). The solid line is the output signal of the low-pass filter 9, and the chain line is the output signal of the low-pass filter 8.

In other words, the signal indicated by the chain line is the blur signal that has passed through the blur mask. Next, when the signal of the low-pass serial filter 8 is subtracted from the signal of the low-pass filter 9 by the operational amplifier 1o, the output signal waveform of the operational amplifier 1o becomes as shown in FIG. 2(d). This is the image contour signal.

The magnitude of this contour signal is appropriately set by the operational amplifier 1o. When this contour signal and the output signal of the low-pass filter 9 are added by the adder 11, the output signal waveform of the adder 11 is obtained as shown in FIG. This is a waveform in which the outline of the image is emphasized.

The contour enhancement algorithm can be expressed mathematically as shown in equation (1).

A-a+F(a-b) ---=----(1)A
: Contour-enhanced signal a; Signal read at normal size (output signal of low-pass filter 9 in Figure 1) b: Blurred signal (output signal of low-pass filter 8 in Figure 1) F: Contour enhancement coefficient (signal given by the operational amplifier 1° in Figure 1) In the waveform diagrams in Figure 2 (,) to (,), waveforms (a), ( The high frequency noise component of b) is completely removed by the low pass filters 8 and 9, and the following waveforms (c) to (-) are drawn.

However, in reality, noise components that cannot be removed by the low-pass filter 8.9 are generated, mixed in the contour signal of waveform (d), and appear in the contour emphasis signal waveform e.

This noise component similarly increases as the degree of contour enhancement increases.

As described above, in the conventional image processing method that performs edge enhancement, the signal read at the normal size is obtained by multiplying the difference signal between the signal read at the normal size and the blur signal by a certain coefficient. Therefore, if a document with noticeable noise is read, even the unnecessary noise will be emphasized and recorded, making the recorded image difficult to see. On the other hand, if an attempt is made to suppress the noise occurring in the document by changing the coefficients, the clear outline will become blurred, and the entire recorded image will become unclear.

The present invention has been made in view of the above problems, and by making the contour enhancement coefficient a function that changes according to the density of the read image, it is possible to perform contour enhancement while suppressing the generation of unnecessary noise, and to The objective is to develop an image processing method that can suppress noticeable noise even when left undisturbed.

The present invention will be described below based on examples and with drawings.

The present invention further considers noise. (1) Since noise does not necessarily exist uniformly in the same magnitude in the amplitude direction of an image signal, it can be processed depending on the image.

■ For humans who observe images, certain parts of the image (
For example, the noise in bright areas is especially noticeable, so you can reduce it.

In short, it has been found that depending on the image and noise content, it is sufficient to reduce the noise in a specific part even if other noises increase somewhat, and this is based on this finding.

In the present invention, in the contour enhancement of equation (1), the contour enhancement coefficient F is given in various ways to perform contour enhancement with noise reduction in a specific portion. The contour enhancement coefficient F may be given any characteristics depending on the image content, but
Representative examples are shown in Figures 3a-h.

In FIG. 3 (-), the contour enhancement coefficient F is unconditionally set to a constant value, which is a conventional example. FIG. 2B shows an example in which a constant value is given to the edge enhancement coefficient F when the absolute value of the signal a or the signal a-b or the value of the signal S exceeds a certain set value. (The dotted line in the figure shows the symmetrical characteristics. The same applies below) As parameters, the signals a, 1a-b
Although the example of l and b is shown, various other methods are possible, such as taking the ratio of signals a and b. The characteristic shown in FIG. 6(b) is a characteristic that emphasizes the outline of areas above the set value in order to prevent noise from increasing where the parameter is below the set value. Following the same idea, (C) in the same figure performs contour enhancement only when the parameter is an intermediate value, (d) in the same figure, ('') t (q)
is a characteristic that increases the degree of edge enhancement as the parameter increases. U) and (h) in the same figure are (→, (
This is the inverse characteristic of q), and therefore, the degree of contour enhancement decreases as the parameter decreases.

FIG. 4 is a block diagram of an apparatus according to an embodiment of the present invention.
This is the main configuration of the device shown in the figure with a part that changes the degree of edge enhancement added. 8,9゜10.11 is the same as in Fig. 1, 12 is a circuit that generates the characteristics shown in Fig. 3 (b) to (g)) with parameters a and b, and 13 is the output signal of circuit 12. This is a circuit (for example, a multiplier) that controls the magnitude of the contour signal (the output signal of the operational amplifier 1o). circuit 13
The output signal enters the adder 11. ' 1 The circuit 12 generates the signals shown in Figure 3 ( This can be easily done using currently commercially available semiconductor devices and known techniques.

In other words, a signal of normal magnitude is passed through the low-pass filter 8 and the high-frequency component noise is removed, and a blurred signal is passed through the low-pass filter 9 and the high-frequency component noise is removed. is subtracted by the operational amplifier 10 and input to the multiplier 13. On the other hand, both signals that have passed through the low-pass filters 8 and 9 are input to a function generator 12, and the output thereof is input to a multiplier 13. At this time, the function generator 12 can take out the signal output in relation to the output of one or both of the filters 8 and 9. The output signal from the multiplier 13 is input to the adder 11, where it is added to the signal passed through the low-pass filter 9 and output as an edge-enhanced signal.

In the embodiments described above, the increase in noise due to edge enhancement has been reduced, but it is clear that when the edge enhancement coefficient F is negative, it weakens the edge and simultaneously reduces noise.

5(-) to 5(e) show examples of waveforms when the contour enhancement coefficient F is changed from negative to positive using the apparatus shown in FIG. The waveform (-) is a diagram in which the contour enhancement coefficient F is given as -1 when the signal a is from 0 to al, and as 1 when the signal a is 0.82 or more between a and a2. Assume that an area where the density of the image is step-like is being scanned, and the waveform of the signal & is obtained as shown in FIG. ” 1 + n 21 n 3 shows the waveform due to noise points included in the image.
) shows the waveform of a-b, and (e) of the same figure shows the waveform of A. As shown in the waveform of , where the value of the contour enhancement coefficient, F, is negative, the noise is small and the contour is weak, where F is 00, it is the same as the waveform of a, and where F is positive, the noise is large. It can be seen that the contours have also become stronger.

As explained above, according to the present invention, the contour enhancement coefficient is changed according to the density of the image, so that the contour is emphasized only in areas with high density of the image, and unnecessary noise is generated in other areas. In addition, it is possible to suppress only the noise occurring in other areas while leaving the necessary outline as is. Therefore, it is possible to partially express the pale or intense feeling of the image, resulting in a recorded image that is easy to see. can get.

[Brief explanation of drawings]

Figure 1 is a block diagram of a conventional contour enhancement device, and Figure 2 (a
) to (e) are waveform diagrams showing the waveforms of each part of the device, and Figure 3 (
, ) to (h) are diagrams showing the characteristics of the contour enhancement coefficient F, and FIG. 4 is a block diagram showing an embodiment of the image processing method of the present invention.
FIGS. 5(-) to 5(e) are waveform diagrams showing waveforms of various parts in the same embodiment. 1... Original, 2... Lens, 3...
...Half mirror-14...Slit, 5...
...Photoelectric converter, 6...Slit, 7...
...Photoelectric converter, 8...Low pass filter, 9...Low pass filter, 10...
... operational amplifier (subtractor), 11... adder,
12... Contour enhancement coefficient generator, 13...
- Multiplier. Name of agent: Patent attorney Toshio Nakao and 1 other person
1'(m Fig. 2 Fig. 3

Claims (4)

[Claims] (1) Subtract the signal obtained by reading the image at a large scanning point from the signal obtained by reading the image at a small scanning point, multiply the signal obtained by the subtraction by an edge enhancement coefficient, and perform the multiplication. The image processing method is characterized in that a signal obtained by reading the image at the small scanning point is added to a signal obtained by reading the image at the small scanning point, and the edge enhancement coefficient is changed according to the density of the image. (2) The image processing method according to claim 1, characterized in that the contour enhancement coefficient is changed according to changes in signals read at small scanning points. (3) The image processing method according to claim 1, wherein the contour enhancement coefficient is changed in accordance with changes in signals read at large scanning points. (4) The image processing method according to claim 1, wherein the contour enhancement coefficient is changed according to changes in signals read at small scanning points and large scanning points.