JPH0636551B2 - Image reading method and apparatus - Google Patents

Description

The present invention relates to an image reading method and apparatus, and more particularly to an image reading method and apparatus capable of preventing the occurrence of moire.

(Prior Art) In a facsimile, a digital copying machine, or the like, one line image sensor including a plurality of photoelectric conversion elements is used as an image reading device for reading a document image. Usually, when a halftone dot photograph of a printed matter is read by a line image sensor in an original used in an office or the like, a striped pattern (moiré) that does not exist in the original image may appear in the output signal. It is known that moire occurs when the original image is a halftone dot photograph and the halftone dot interval is close to the sampling interval of the image sensor. As a method of suppressing such moire, for example, a halftone dot frequency is uniformly removed by a low pass filter.

(Problems to be Solved by the Invention) As described above, when smoothing is performed uniformly using a low-pass filter, there is a problem in that the edges are blurred and the resolution is reduced.

The present invention has been made in view of such points,
An object of the invention is to realize an image reading method and apparatus capable of removing moire without lowering the resolution.

(Means for Solving the Problems) A first invention for solving the above problems is to read a document image at at least two different sampling intervals to obtain a first signal, and each of the first signals. To obtain a second signal having a waveform having a spatial frequency part corresponding to the image and a spatial frequency part corresponding to the reading sampling interval, and processing them to correspond to the reading sampling interval. And a step of obtaining a fourth signal by performing an inverse orthogonal function transformation on the third signal and a step of obtaining a fourth signal by substantially removing the above portion.

A second aspect of the present invention provides at least two original images.
Means for reading at different sampling intervals to obtain a first signal, and each of the first signals being subjected to orthogonal function conversion to obtain a waveform having a spatial frequency portion corresponding to the image and a spatial frequency portion corresponding to the reading sampling interval. Means for obtaining the second signal, means for processing each of the second signals to obtain a third signal in which a portion corresponding to the read sampling interval is substantially removed, and for the third signal And an inverse orthogonal function conversion to obtain a fourth signal.

(Operation) The orthogonal transform has a feature that in the transform result, the energy distribution is mainly concentrated in the low frequency region, and the components in which the edge and line information is reflected are concentrated in the high frequency region.

Therefore, when a document is read at different sampling intervals and subjected to orthogonal transformation (for example, Fourier transformation), small peaks of components including moire appear in the high frequency region, and their positions differ depending on the sampling intervals. On the other hand, the energy distributions in the low frequency range (near zero frequency) are almost the same.

The present invention takes advantage of this fact and links the deviation of the sampling interval of reading with the deviation of the frequency position of the small peak portion including the moire on the orthogonal transformation plane, and the signals read at the different sampling intervals. They are processed to suppress moire without lowering the resolution.

That is, for example, by comparing the absolute values of the respective converted signals for each spatial frequency, appropriately selecting the smaller one, and combining (combining) for each frequency, the portion with less moire is automatically detected. Selected by this
Moire can be suppressed.

Further, in the method of removing the moire component by the filter, the edge information of the document, which is minute but valuable, which is superposed on the same frequency range at the same time as the moire component, is completely removed, and the faithful reproduction cannot be performed. . On the other hand, in the present invention, since the method of selecting the read signal that has not been subjected to any original processing is adopted, the useful information contained in the read signal is utilized as it is.

(Example) Hereinafter, an example of the present invention will be described in detail with reference to the drawings.

FIG. 1 is a flow chart showing an embodiment of the method of the present invention. The present invention will be described in detail below with reference to this flowchart.

Although the signal processing of the present invention is actually performed in the two-dimensional direction, it will be described as one-dimensional signal processing in order to simplify the description.

FIG. 2A is a diagram in which the reflectance is obtained by scanning the halftone image. The horizontal axis represents the distance x, and the vertical axis represents the reflectance r (x). In the case of a halftone image, its reflectance r (x) is discrete as shown in the figure. T is the halftone image period. By the way, investigating what kind of frequency component an image signal is composed of is an important means in signal processing and analysis. A so-called orthogonal function transformation is used to examine the frequency components. Here, a unitary matrix (whose inverse matrix is equal to a transposed matrix) A (M), A (N) is applied to an image x of size M × N pixels Transformation of the form is called two-dimensional orthogonal transformation. The orthogonal transformation has features such as concentration of energy distribution in low-frequency components and reflection of edge / line information in high-frequency components as a result of transformation, and is used when extracting features from an image. In the case of orthogonal transformation, it is common to handle it including discrete Fourier transform, and discrete Fourier transform is often used to represent orthogonal transform.

For example, given a two-dimensional image, what frequency (called spatial frequency) component it is composed of is investigated. When the Fourier transform is performed on the halftone dot image r (x) shown in FIG. 2 (a) and it is displayed in the spatial frequency plane, it becomes as shown in FIG. 2 (b). The horizontal axis is the spatial frequency ν, and the vertical axis is the Fourier transform result R (ν)
Indicates. R (ν) is expressed by the following equation.

R (ν) = f {r (x)} (1) Here, f Fourier transform is shown.

Step Read the original image at least two different sampling intervals.

FIG. 3A shows a function g ₁ (x) of the sampling interval d ₁ by the first image sensor, and FIG. 4A shows a function g ₂ (of the sampling interval d ₂ by the second image sensor. x). In this way, the same original image is sampled at two different sampling intervals d ₁ and d ₂ , and the image information is read. As the means for reading the image information, for example, the above-mentioned line image sensor is used.

Step The read image is transformed by orthogonal function.

When the sampling function g ₁ (x) shown in FIG. 3 (a) is Fourier transformed and displayed in the spatial frequency plane, a Fourier transform image G ₁ (ν) as shown in FIG. 3 (b) is obtained. Here, G ₁ (ν) is represented by the following equation.

G ₁ (ν) = f {g ₁ (x)} (2) FIG. 3 (c) is a halftone dot image r (x) shown in FIG. 2 (a).
Is an image obtained by sampling with a sampling function g ₁ (x) shown in FIG.
₁ (x) is represented by the following equation.

h ₁ (x) = r (x) g ₁ (x) ... (3) When the sampled image h ₁ (x) represented by the equation (3) is Fourier transformed and displayed in the spatial frequency plane, the Fourier transformed image H ₁ (ν) is as shown in FIG.
Here, H ₁ (ν) is represented by the following equation.

Here, the symbol “*” indicates a convolution operation.

Next, similar processing is performed for FIG. When Fourier transform is applied to the sampling function g ₂ (x) shown in FIG. 4 (a), a Fourier transform image G ₂ (ν) as shown in FIG. 4 (b) is obtained in the spatial frequency ν region. Where G
₂ (ν) is expressed by the following equation.

G ₂ (ν) = f {g ₂ (x)} (5) Next, the halftone image r (x) shown in FIG. 2 (a) is sampled by the sampling function g ₂ (shown in FIG. 4 (a). x) sampled image h as shown in FIG.
₂ (x) is obtained. This sampled image h ₂ (x)
Is expressed by the following equation.

h ₂ (x) = r (x) · g ₂ (x) ... (6) Next, when the Fourier transform is performed on the sampling image represented by the equation (6) and the spatial frequency (ν) plane is displayed, the Fourier transform image H ₂ (ν) is as shown in FIG. 4 (d). Here, H ₂ (ν) is represented by the following equation.

The Fourier transform described above is a continuous Fourier transform.
Actual signal processing is performed discretely. Sampling images h ₁ (x) and h ₂ (x) shown in (c) of FIGS. 3 and 4
When the Fourier transform is discretely performed on the Fourier transform image H as shown in FIGS.
₁ (n) and H ₂ (n) are obtained. Here, n on the horizontal axis is a discrete spatial frequency. H ₁ (n) and H ₂ (n) are represented by the following equations, respectively.

H ₁ (n) = d {h ₁ (x)} ... (8) H ₂ (n) = d {h ₂ (x)} ... (9) where the symbol “d” is a discrete Fourier transform. .

Steps The signals subjected to the orthogonal function conversion are compared to remove unnecessary signal components.

As is apparent by comparing and comparing the waveforms shown in FIGS. 5 (a) and 5 (b), there are two small peaks indicating moire on both sides of the peak centered at the zero frequency. Moreover, the positions of these peaks are different between FIGS. 5 (a) and 5 (b).
What we want to find here is a peak-only signal centered on a zero frequency with no small peaks on either side. It seems that it would be better to apply a low-pass filter to the waveforms shown in Figs. 5 (a) and 5 (b) to remove the small peaks on both sides. However, since the original image itself also includes frequency components at the positions of the small peaks on both sides, it is impossible to faithfully reproduce the original signal.

Therefore, in the present invention, the absolute values of the two discrete Fourier transform images H ₁ (n) and H ₂ (n) are compared, and the smaller one is newly extracted for each spatial frequency. (N). H (n) is represented by the following equation.

H (n) = cmin {H ₁ (n), H ₂ (n)} (10) where cmin {} is a function that takes the smaller absolute value of the complex number. FIG. 5C is a diagram showing the waveform of H (n). It can be seen that the small peaks on both sides are almost completely removed.

Step The signal after removing unnecessary signal components is inversely transformed by the orthogonal function to be returned to the read image signal.

Signal H after removal of unnecessary signal components shown in FIG.
When discrete inverse Fourier transform is applied to (n), an image signal h (x) without moire as shown in FIG. 5D is obtained. Here, h (x) can be expressed by the following equation. Further, p (x) is a function obtained by interpolating h (x).

h (x) = d ⁻¹ {H (n)} (11) The image shown in FIG. 5 (e) is the Fourier transform of the image signal h (x) shown in FIG. The moiré-free waveform shown in Fig. 3C is repeated at 1 / d ₁ .

FIG. 6 is a configuration block diagram showing an embodiment of the device of the present invention. In the figure, reference numerals 1 and 2 denote first and second image sensors that receive an image input. The original image is, for example, the 7th
Image sensor 1, via a reading optical system as shown in the figure
Entered in 2. The optical image information from the original 31 is mirror 3
2, the light enters the half mirror 34 through the lens 33, the half mirror 34 divides the optical image information into two, and the transmitted light is the first light.
Of the image sensor 1 and the reflected light enters the second image sensor 2.

The image sensors 1 and 2 need to have different sampling intervals as described above, and the mounting pitches of the sensor elements (photoelectric conversion elements) are d ₁ and d ₂ , respectively.
As the image sensors 1 and 2, a one-dimensional line image sensor may be used for scanning, or a two-dimensional image sensor may be used for reading at once.
As a condition, it is necessary that the distances d ₁ and d ₂ between the sensor elements are different, and in the case of a two-dimensional image sensor,
It is preferable that the vertical and horizontal intervals are different from each other.
Further, it is preferable that the interval of one image sensor is not an integral multiple of the interval of the other image sensor, and the ratio of the pitch intervals d ₁ and d _{2 of} both image sensors is 1 to 2 times, especially 1. It is desirable that it is 1 to 1.8 times.

Output signals from the two image sensors 1 and 2 are converted into digital data by A / D converters 3 and 4, respectively, and stored in memories 5 and 6. The image data stored in the memories 5 and 6 are sequentially read and subjected to discrete Fourier transform by the discrete Fourier transform circuits 7 and 8, respectively.
It is stored in the memories 9 and 10. The sections in which the discrete Fourier transform of the two image sensors 1 and 2 are performed are preferably the same on the image of the original, and therefore the ratio of the sample numbers of the image sensor 1 and the image sensor 2 is d ₂ : d.
_1, i.e., it is preferable that the reversed ratio of the pitch _d 1, _{d 2.}

The Fourier transform data stored in the memories 9 and 10 are sequentially read out, converted into absolute values by the absolute value circuits 11 and 12, and then compared by the comparison circuit 13 for each spatial frequency. The comparison circuit 13 gives a changeover signal to the changeover switch 14 so as to select one having a smaller absolute value according to the comparison result. The changeover switch 14 changes over to the data side having the smaller absolute value of the Fourier transform data stored in the memories 9 and 10, and selects the data having the smaller absolute value and sends it to the memory 15. The data having the smaller absolute value thus selected is sequentially stored in the memory 15.

The data stored in the memory 15 is sequentially read, inverse Fourier transformed by the discrete inverse Fourier transform circuit 16, and returned to the image read signal. This image reading signal is one from which moire has been removed. The signal data returned to the image reading signal by the discrete inverse Fourier transform circuit 16 is stored in the memory 17. The image read signal data stored in the memory 17 is read out as needed, output-displayed on a CRT, or printed out by a printer. The image output and displayed in this manner is a high-quality image without moire.

The experimental results of the present invention will be described below. 100 cotton / inch 45 ° halftone image with 200 pixels / inch and 15
Aperture 100 with two sampling intervals of 0 pixel / inch
When a 2 inch × 2 inch area was read by a drum type scanner of μm, 42-dot / inch and 7-dot / inch original dot patterns, which were not present in the dot image of the original, were generated. Therefore, these two read images (400 pixels x 40
Two-dimensional discrete Fourier transform is performed on the 0 pixel and 300 pixel × 300 pixel by a computer, and the low spatial frequency component (300 × 3) is included in the 400 × 400 Fourier plane image.
00) and a 300 × 300 Fourier plane image, a component having a smaller absolute value is selected for each spatial frequency component, and a 300 × 300 Fourier plane image is newly created. An image obtained by subjecting this image to a two-dimensional discrete inverse Fourier transform was a good halftone image without moire.

In the above description, the case where the original image is read at two different sampling intervals is taken as an example, but the original image may be read at three or more different sampling intervals. Further, in the above description, the case where one-dimensional signal processing is performed has been described as an example, but if the signal processing according to the present invention is performed in the two-dimensional direction, the moire removing effect is further improved. Further, in the above description, the case where the orthogonal function conversion is performed on the Fourier transform plane has been described as an example, but the present invention is not limited to this, and is applicable to all conversions using an orthogonal function. For example, it can be applied to Walsh-Hadamard conversion or the like.

Further, in the above description, the calculation is performed by taking the smaller absolute value of the Fourier transform images to be compared in order to remove the moire, but the present invention is not limited to this, and any operation that can remove the peak due to moire is possible. Any method may be used. Furthermore, a fast algorithm for discrete Fourier transform / inverse transform, for example, FFT (Fast Fourier Transform)
And WFTA (Winograd Fourier Transform Algorithm)
Using m) etc., the calculation time of the computer can be shortened.

(Effect of the invention) As described in detail above, according to the present invention, by orthogonally transforming respective signals read at different sampling intervals, and processing the respective signals obtained by the transformation,
That is, for example, the moire component can be suppressed without spoiling useful information by performing the process of selecting and combining the smaller absolute value for each spatial frequency, and a high-quality image can be obtained. It is possible to realize an image reading method and device capable of reproducing the image. Therefore, the practical effect is extremely large.

[Brief description of drawings]

FIG. 1 is a flow chart showing an embodiment of the method of the present invention,
2 to 5 are explanatory views of the method of the present invention, FIG. 6 is a configuration block diagram showing an embodiment of the apparatus of the present invention, and FIG. 7 is a diagram showing an example of the configuration of a reading optical system. 1, 2 ... Image sensor 3, 4 ... A / D converter 5, 6, 9, 10, 15, 17, ... Memory 7, 8 ... Discrete Fourier transform circuit 11, 12 ... Absolute value circuit 13 ...... Comparison circuit, 14 ...... Changeover switch 16 ...... Discrete inverse Fourier transform circuit 31 ...... Document, 32 ...... Mirror 33 ...... Lens, 34 ...... Half mirror

Claims (2)

[Claims] 1. A step of reading a document image at least at two different sampling intervals to obtain a first signal, and an orthogonal function conversion of each of the first signals, and a spatial frequency portion corresponding to the image and a reading sampling interval. To obtain a second signal having a waveform having a spatial frequency portion corresponding to the third signal, and processing each of the second signals to obtain a third signal in which a portion corresponding to the read sampling interval is substantially removed. An image reading method comprising: a step of obtaining and a step of performing an inverse orthogonal function transformation on the third signal to obtain a fourth signal. 2. A means for obtaining a first signal by reading an original image at at least two different sampling intervals, and an orthogonal function conversion of each of the first signals to obtain a spatial frequency portion corresponding to the image and a reading sampling interval. Means for obtaining a second signal having a waveform having a spatial frequency portion corresponding to, and a third signal obtained by processing each of the second signals to substantially eliminate the portion corresponding to the read sampling interval. An image reading apparatus comprising: obtaining means and means for performing an inverse orthogonal function transformation on the third signal to obtain a fourth signal.