JPH065879B2 - Image processing device - Google Patents

Description

TECHNICAL FIELD The present invention relates to a digital copying apparatus that handles an image as an electric signal,
The present invention relates to an image processing device such as a facsimile device.

<Prior Art> Generally, an image is sampled by a CCD sensor or the like,
A so-called digital copying apparatus that outputs digitalized data from a digital printer such as a laser beam printer to reproduce an image is becoming widely used instead of a conventional analog copying apparatus due to the development of digital equipment. Since such a digital copying apparatus reproduces a halftone image, it is common to perform gradation reproduction by a dither method or a density pattern method. However, this method has the following two major problems.

(1) When the original is a halftone dot image, a periodic striped pattern which is not present in the original appears in the copied image.

(2) If the original contains line drawings, characters, etc., the edges are cut off by the dither processing, and the image quality is degraded.

The phenomenon of (1) is called a moire phenomenon, and its cause is considered to be A) a halftone original and moire due to input sampling B) a halftone original and moire due to a dither threshold matrix. The phenomenon of A) is that the halftone dot frequency f _o (= 1 / P _o ) [PE is determined by the halftone dot pitch P _o [mm] of the halftone dot original.
L / mm] n times higher frequency nf _o [PEL / mm] and the input sampling frequency f _s (1 / P _s ) [PEL / mm] obtained from the input sensor pitch P _s [mm] Δf = | F _s −nf _o | [PED / mm] (1) A beat frequency is generated, which causes moire.

In the phenomenon of (B), the dither threshold is generally fattin.
When arranged in a dot-concentrated type such as g-type, the output image also has a pseudo halftone dot structure, which causes a beat between the input halftone dot original and a moire phenomenon. When the repetition cycle pitch of the dither threshold is P _D [mm] on the recording paper, the spatial frequency is f _D = 1 / P _D [PEL / mm], and the beat frequency is Δf = | f _o −f _D | [PED / mm] (2) is most noticeable.

The strongest occurrence of the above two moire phenomena is in (B). This is because, in the phenomenon of (A), generally, n is 3 to 6 as n at a frequency n times higher than that of a halftone dot original, and the transfer function (MTF) of the optical system or the like leading to the sensor is considerably lowered at that frequency. The contrast of moire fringes is also low.

The moire phenomenon caused by such a cause significantly deteriorates the quality of the output image. For this reason, various measures and studies have been made since ancient times. For example, the method using the random dither method can remove moire, but it produces grain-like graininess and does not deteriorate image quality. J. Opt. Soc. Am. , Vol. 6
6, No10, October 1976 P985, Paul G. et al. A proposed by Roetling
RIES compares the average values of the densities before and after binarization and feeds back the threshold value so that they become equal. However, this method is complicated in hardware and has a sufficient moire removing effect. Not.

On the other hand, the Institute of Image Electronics Engineers, Yohka 83-3 P13 "Dotization of mixed text / photo image" The fine dot method found in Takashima et al. Is a method of blurring a dot image (or averaging up to surrounding pixels) Moire is removed because the pattern is re-dotted, and there is little granular noise.

However, blurring (or averaging with surrounding pixels) inevitably reduces the resolution. That is, if the moiré is removed, the resolution is lowered, and if the resolution is kept, the moiré is not removed. Therefore, it is indispensable to extract only the halftone image area in advance and apply the method to only that portion. Therefore, so-called image area separation technology is required. This image area separation technology is difficult to realize in the current level because it is difficult to obtain a method with high accuracy and high speed, especially a method suitable for hardware. Moreover, even if an image area separation technique were obtained, such a method would not be sufficiently satisfactory because even high frequency components in the image would be averaged and smoothed.

On the other hand, with respect to the problem (2), the character / line drawing of the document is shredded by performing the dither process, and the quality of the print is deteriorated especially because the dedge portion is cut off. This phenomenon is particularly remarkable when the dither pattern is the dot concentrated type such as the aforementioned Fatting type.

<Purpose> An object of the present invention is to provide an image processing apparatus which can eliminate the above-mentioned two secondary points and reproduce an image with high quality and high definition. That is, in the present invention, the moire phenomenon that occurs during halftone dot originals is removed, and it is possible to prevent the characters and line images from being shredded into small pieces, and to prevent the high frequency components in the image area from decreasing. Was made. Further, the method can be realized with a simple circuit configuration and can be provided at a low cost.

<Embodiment> (Basic Structure) FIGS. 1 to 6 FIG. 1 shows the basic structure of an image processing apparatus according to this embodiment. The image processing apparatus includes an edge detector a, an edge enhancer b, a smoothing device c, and a mixer d. The edge detector a detects the edges of characters, line drawings, and images as described later, and the halftone dots of the halftone dot image have a spatial frequency characteristic that is not detected as an edge. The edge enhancer b outputs an original image or an edge-enhanced image signal in which the original image and edges are mixed at a certain ratio. The smoothing device c smoothes the image. The mixer d changes the mixing ratio of the edge-enhanced image and the smoothed image according to the signal from the edge detector and outputs the image. In this way, the halftone dots of the halftone dot image are determined to be non-edge regions, and smoothing is performed by smoothing to prevent moire. Further, the edges of characters, line drawings, and images are determined to be edge regions, and the edges are emphasized to prevent the characters from becoming halftone dots and to reduce the sharpness of the images.
Further, since the edge area and the non-edge area are continuously connected, the texture does not change at the boundary.

Next, the principle of this embodiment will be described from frequency characteristics.
First, the screen ruling of the halftone image of the original is usually black and white 1
20 to 150 lines, 133 to 175 lines in color. And the screen angle is 0 when moire is likely to occur.
It is at 45 degrees. The main scanning direction halftone dot pitch during line reading is maximum at 45 degrees and low in spatial frequency, and minimum at 0 degrees and high in spatial frequency. Table 1 shows the spatial frequencies when the screen angle is 0 degree and 45 degrees.

The frequency characteristic of such a halftone image has peaks at the fundamental frequency and its high frequency as shown in FIG. Also a character image,
The frequency characteristics of continuous-tone photographic images are shown in Figs. 2b and 2c, respectively.
become that way. For such a mixed image of characters, photographs, and halftone dots, the spatial filter of the edge detector, edge enhancer, and smoothing device of the present embodiment has frequency characteristics satisfying the following conditions.

Condition 1. The peak frequency of the spatial filter of the edge detector is lower than the first harmonic frequency of the halftone dot image.

Condition 2. The peak frequency of the spatial filter of the edge enhancer is higher than that of the spatial filter of the edge detector.

Condition 3. The frequency characteristic of the spatial filter of the smoothing device is sufficiently lowered at the first harmonic frequency of the halftone image.
Also, it is sufficiently reduced at the frequency corresponding to the output dither cycle.

Although there are various spatial filters for edge detection, if the matrix size that affects the scale of the hard circuit is fixed, the first derivative filter has a peak at a lower frequency than the second derivative filter. However, the second derivative filter has no directionality, but the first derivative filter has directionality, and the square root of the sum of the squares of the slopes of at least two directions, or an approximate expression thereof, is the absolute slope of at least two directions. It is necessary to take the sum of the values or the maximum absolute value of the inclination in at least two directions. Further, the first derivative is more resistant to point noise than the second derivative. As described above, the first-order differential filter is preferable as the spatial filter of the edge detector a.

Further, as the spatial filter of the edge enhancer b, the second order differential filter having no directivity and having a peak at a higher frequency is superior to the first order differential filter.

For the sake of simplicity, the results of the one-dimensional fast Fourier transform (FFT) calculation of the relationship of the frequency characteristics of the above various spatial filters will be shown. As an example, the reading sampling interval of the input system is 1/16 mm, and the output system is 16 dots / mm, 4 x 4
The calculation is performed using the dither matrix of. If the period of the dither pattern is converted to the spatial frequency, 4
It is 1 / mm. Also, in the case of reading with 1/16 mm sampling, only the frequency of 81 / mm can be detected by the sampling theorem.

When the matrix size is 5 × 5, the one-dimensional FFT of the second derivative filter (-1, 0, 2, 0, -1) is shown in FIG. 3 and the first derivative filter (-1, 0, 0, 0, 1 of 1)
The dimensional FFT is shown in FIG. 4, and another one-dimensional differential filter (-
A one-dimensional FFT of (1, -1, 0, 1, 1) is shown in FIG.

The peak positions are 4 1 / mm, 2 1 / mm, 2.
It is 5 1 / mm. Comparing this with the spatial frequency of the halftone dot image in Table 1, Condition 1 is satisfied for all the number of lines in Table 1 in the first-order differential filter, but 45 in 120 lines and 133 lines in the second-order differential filter. Condition 1 is not satisfied at °, and halftone dots are detected as edges. When comparing two types of first-order differential filters, the pulse width was increased (-
1, -1,0,1,1) is superior. This is because the intensity of the second peak becomes smaller when the pulse width is increased, and the edge region (edge emphasis is applied to this region) can be detected broader when the pulse width is increased. Edge detection is performed by the first derivative filter (-1, -1,
0, 1, 1) and the edge enhancement is a second derivative filter (-1, 0, 2, 0, -1), the peak frequencies are 2.5 1 / mm and 4 1 / mm, respectively. Meet That is, a region in which edge enhancement is wider than edge detection is extracted, and a spatial filter that produces sharp edges is used in edge enhancement.

Next, a 5 × 5 smoothing filter (1, 1, 1,
A one-dimensional FFT of (1, 1) is shown in FIG. 120 line 45
The intensity becomes small at the fundamental frequency of the halftone dot image of ° or more and 3.341 1 / mm or more. Also, the condition 3 is satisfied because the strength is sufficiently small at 4 × 4, the pitch of the dither matrix, 41 / mm.

In this embodiment, the edge detector, the edge enhancer, and the smoothing device use the spatial filters having the frequency characteristics as in the above-mentioned conditions 1 to 3, so that the flat portion of the image and the halftone dot image are determined to be non-edge regions and smoothed. Are averaged, and the edge portion of the character, line drawing, and image is determined as an edge region, and the edge is emphasized. The boundary between the edge region and the non-edge region is continuously connected by changing the mixing ratio in the mixer according to the signal from the edge detector. As described above, moire in a halftone image is prevented, halftone dots are prevented, and the sharpness of the image is prevented from decreasing, and discontinuous texture changes between edge regions and non-edge regions do not occur. Further, since the spatial filter does not need to have a large matrix size, the scale of the hardware circuit can be reduced, which is also advantageous for LSI implementation.

FIG. 7 is a block diagram showing an embodiment of the present invention, S1 is an input image signal, 1 is a differential value detecting section for detecting the absolute value of the primary differential value of the input image signal S1, and corresponds to FIG. 1a. . S2
Is a differential signal connected to the output of the differential value detecting section 1, 2 is a control signal generator for generating control signals S3 and S4 from the differential signal S2, S3 is an output of the control signal generator 2, and a control signal S4
Is a control signal complementary to the control signal S3 at the output of the control signal generator 3, and 3 is a smoothing processing unit for smoothing the input image signal S1 and corresponds to FIG. 1C. S6 is a smoothed image signal smoothed by the smoothing processing unit 3, 4 is a multiplier that takes the arithmetic product of the smoothed image signal S6 and the control signal S3,
S7 is the output of the multiplier 4, 5 is an edge enhancing unit for enhancing the edge portion of the input image signal S1, S8 is an edge signal of the edge enhancing unit 5, S9 is a constant given from the outside, 6 is an edge signal S8 and a constant S9. Multiplier S1 that takes the arithmetic product of
0 is the edge signal output from the multiplier 6, 7 is an adder that takes the arithmetic sum of the edge signal S10 and the input image signal S1, and the edge emphasis unit 5, the multiplier 6 and the adder 7 are the edges of FIG. Construct the highlighter b.

S11 is an edge-enhanced image signal that is the output of the adder 7, 8 is a multiplier that calculates the arithmetic product of the edge-enhanced image signal S11 and the control signal S4, and S12 is the output of the multiplier 8.
9 is an adder for taking the arithmetic sum of the outputs S7 and S12, S1
The output 3 of the adder 9 is a processed image signal. Here, the multipliers 4 and 8 and the adder 9 constitute the mixer d of FIG.

In the edge enhancing unit 302 corresponding to the edge enhancing unit b, the multiplier 6 multiplies the output of the edge amount (edge detect) detecting unit 5 and the control signal S9 described above. It goes without saying that the multiplier 6 can be composed of a ROM or the like, and the control signal S9 may be a coded signal instead of the multiplication coefficient itself.

When the pixel of interest in the image processing is A, the adder 7 adds the pixel of interest and a value multiplied by a coefficient having edge detection, and the pixel of interest is edge-emphasized. Mixing unit 305
In, the output of the edge enhancement unit 302 and the smoothing process 3
Is a circuit unit that mixes the outputs of the control signals in a proper ratio, according to the output of the differential value detecting unit 1 input to the control signal generating unit 2 in the previous stage.
S3 and S4 are output. As will be described later, S3 and S4 are complementary control signals, but they are not necessarily limited. S
The characteristics of 3 and S4 can be freely selected and set by the control signal S5. In the multiplier 8, the output of the edge enhancement unit performs the multiplication determined by S4, in the multiplier 4, the smoothing processing unit performs the multiplication determined by S3, and the outputs of the multipliers 4 and 8 are mutually added by the adder 9. They are added and this becomes the image processing output.

The block diagram of FIG. 7 can also be expressed as an equation.

First, the differential value detection unit 1 and the control signal generator 2 have the following formula 1
Is calculated.

Here, I is the input image data and E is the control signal S4.
Is a function for normalizing the control signal S4 to a maximum value of 1 and performing gamma conversion.

The following output is obtained from the adder 7.

G is the output of the adder 7, and k ₁ is the constant of the constant signal S9. Then, the following output is obtained from the smoothing processing unit 3.

H is an output of the smoothing processing unit 3. Therefore, the value Out of the output S13 of the adder 9 can be expressed by the following equation 2.

The inside of [] shown in the above equation is a kernel that convolves with the image signal I. Various modifications can be considered for the kernels * 1 to * 4, and an example thereof is shown in Table 2.

FIG. 8 shows an example of the characteristic of the function of the control signal generator 2.

At E = f (x), E = 0 for 0x <0.2 E = 1.67x−0.33 for 0.2x <0.8 E = 1 for 0.8x1. However, the input / output signals will be standardized to 0 to 1 for description.

FIG. 9 and FIG. 10 show the edge detector 1 by the first derivative.
It is a diagram for explaining the operation of (hereinafter referred to as a differential value detection unit), and is shown in one-dimensional main scanning.

Since the differential value detecting unit 1 is a kind of band pass filter as already described, the input image signal S1 such as a halftone image having high frequency components shown in FIG. , -1, 0, 1, 1) is convolved in the main scanning direction, the output signal S2 becomes a small value such as 0.1-0.2.

On the other hand, in the case of an input image signal of relatively low frequency (for example, a vertical line of a character), the output signal S2 has a large value due to the same convolution as shown in FIG.

Here, the control signal generator 2 for gamma exchange sets the control signal S3 to 1 and the control signal S4 to 0 when the differential signal S2 is smaller than 0.2 as shown in FIG. When the differential signal S2 is larger than 0.8, the control signal S3 is set to 0,
The control signal S4 is set to 1. Furthermore, the differential signal S2 is 0.2
Between 0.8 and 0.8, as shown in FIG. 2, the differential signal S is always set so that the sum of the control signal S3 and the control signal S4 becomes 1.
It changes according to 2.

On the other hand, the input image signal S1 is connected to the input of the differential value detecting unit 1 and also to the inputs of the smoothing processing unit 3 and the edge detecting unit 5 at the same time.

FIG. 11 shows the operation of the smoothing processing unit 3, and shows a one-dimensional example in the main scanning direction for the sake of explanation. In this case, the kernel is (1,1,1,1,1,) and all contents are 1
And is a low-pass filter configured to output the average of 5 pixels, and the input image signal S1 becomes the smoothed image signal S6.

Further, FIG. 12 shows the operation of the edge emphasizing section 5, and again shows one dimension in the main scanning direction for the sake of explanation. The kernel is (-1,0,2,0, -1) and has the well-known second-order differential edge detection characteristic, and the output S8 has 0 at the flat portion and positive and negative peaks at the edge portion. .

Further, the edge signal S8 is multiplied by a constant S9 by the multiplier 6, and added with the image input signal S1 by the adder 7 to become an edge emphasis signal S11. Although not shown, since the edge signal S10 is slightly delayed with respect to the input image signal S1, a delay circuit for timing the edge signal S10 input to the adder 7 and the input image signal S1 is actually provided. To be

By the way, the output of the differential value detecting unit 1 is large, that is, the control signal S3 is small and S4 is large at the edge portion. Conversely, when the differential signal S2 is small, S3 is large and S4 is small. Further, as described in FIG. 8, S3 and S4 are gamma-converted so that the sum is always 1. Therefore, the multipliers 4 and 8
The sum of the outputs is the edge emphasis signal S when the differential signal S2 is large.
There are many 11 components, and when S2 is small, the smoothing signal S6 is controlled to have many components.

FIG. 13 shows this state, and it is shown that the differential signal S2 detects an edge in a portion of the input image signal S1 excluding a short cycle vibration (corresponding to a halftone dot cycle).

The control signal S4 is gamma-converted of the differential signal S2, and the signal S2 other than the four peaks shown in FIG. 13 is set to zero. Naturally, S3 is 1-S4. Further, FIG. 13 also shows the smoothing signal S6 and the edge emphasis signal S11. In FIG. 13, a processed signal S13 is a signal in which S6 and S11 are added at a ratio of S3 and S4 to smooth the halftone dot portion and emphasize only the edge portion.

Next, each block in FIG. 7 will be described in detail.

(Differential Value Detection Unit 1) FIG. 14 is a detailed circuit diagram of the differential value detection unit 1.

The 5-line buffer 301 shown in FIG.
The output of is input.

In the figure, 306 is a primary differentiator whose output 306
-A is a data part 306-c and a sine part 3 indicating positive / negative.
06-b, the sign unit 306-b is input to the select signal of the selector 308, and the inverter 307
By selecting either +,-inverted data or 306-c data by
8-a is obtained. Similarly, the absolute value of the output of the primary differentiator 312 is output from the selector 311, and the adder 309 outputs the absolute value.
Thus, 308-a and 311-a are added, and the sum of the two-direction primary differential values is output from the adder 309.

Next, FIG. 15 is a block diagram showing in detail the first-order differentiators 306 and 312 shown in FIG.

First, the block X in FIG. 15 will be described in order to explain the basic operation of the primary differentiating circuit.

First, all shift registers in FIG. 15 are shifted in synchronization with an image transfer clock not shown in the figure. For simplicity of explanation, all the multiplication coefficients of the multipliers 243 to 247 are set to 1. According to the timing chart of FIG. 16, the output of the shift register 230 at t-3 is Sn, m.
−1 + Sn, m−2, and the output of the shift register 231 at t−2 is Sn, m + Sn, m−1 + Sn, m.
-2, and the output of the shift register 232 at t-1 is Sn, m + 1 + Sn, m-1 + Sn, m-2, at t
The output of the adder 260 at 0 is Sn, m + 2 + Sn,
m + 1 + Sn, m + Sn, m-1 + Sn, m-2. In this way, the added value of 5 pixels in the main scanning direction is calculated in the block X. Here, by setting the multiplication coefficients of the multipliers 243-247 to a, b, c, d, and e, the output of the adder 260 is e · Sn, m + 2 + d · Sn, m +.
1 + c · Sn, m + b · Sn, m−1 + a · Sn, m−
It becomes 2.

Similarly, it can be seen that the circuits after the shift register 223 and the circuits after the shift register 233 operate.

By the way, when the first-order differential to be obtained is expressed by * 1 and * 2 in the equation 1, when the pixel of interest is the n-th line, the kernel elements of the n-2 line and the n-1 line are the same, and the n + 1 line is And n + 2 line kernel elements are equal. Therefore, in the image data, the n-2 line and the n-1 line are added 23
The circuit scale can be reduced to ½ by adding 1 and then performing the first-order differential processing such as equations 1 * 1 and * 2. Also, n
The same applies to the + 1st line and the n + 2th line.
In this way, the adder 400 can obtain the added value corresponding to the car name for five lines.

In the case where the multipliers 238 to 252 have a simple multiplication value of 1or-1or0, as shown in FIG. 17, they can be easily configured by the inverter 291 and the selector 292. It can be switched to 1 or -1 and set to 0 by CL.

Further, in order to obtain the first derivative of the kernel * 1 in Expression 1, the multiplication coefficients of the multipliers 238 to 242 are set to 1 and the multipliers 243 to 2
The multiplication coefficient of 47 is 0, and the multiplication coefficients of the multipliers 248 to 252 are -1.

To obtain the first derivative of the kernel * 2 in Equation 1, the multiplier 2
42, 238, 239, 240, 241 have multiplication coefficients of 1, 1, 0, -1, -1, and multipliers 243 to 247 have multiplication coefficients of 1, 1, 0, -1, -1. 248
The multiplication coefficients of ˜252 are 1,1,0, −1, −1.

Also, the circuit configuration for calculating the first derivative such as * 1 and * 2 in Table 2 can be realized by FIG. 18, but since the operating principle is the same as in FIG. 15, how to give the coefficient to the multiplier is Omit it.

(Edge Enhancement Unit 5) Next, the edge enhancement unit 5 is shown in FIG.

As shown in FIG. 20, the input image signal S1 shown in FIG. 7 is composed of data for five consecutive lines of image data, and the input image data input to the 5-line buffer 301 is stored in the 5-line buffer. After being processed, the image data is simultaneously output for five lines and is output pixel by pixel in the main scanning direction of the image data in synchronization with an image transfer clock (not shown).

Further, as shown in FIG. 21, consider the peripheral image data when the attention area S of the image area is further expanded and the attention pixel data is Sn, m. Image data for 3 lines of the n-2th line, n, n + 2th line is input to the edge emphasis unit 5 of FIG. 19 among the input image data S2, and the target pixel of image output is set to Sn, m. .

In the figure, 201 to 211 are 1-bit shift registers. The image data S1 is stored in the shift registers 201 to 20.
3, 204 to 208, 209 to 211 shift in synchronization with an image transfer clock (not shown). This timing chart is shown in FIG. In FIG. 22, the output of the shift register at a certain timing T is shown in () in FIG.

The adder 213 is the shift registers 203, 204, 20.
8, 211 output data Sn-2, m, Sn, m-2,
Sn, m + 2, Sn + 2, m are added, and the added data is multiplied by -1 in the multiplier 214. The pixel of interest Sn, m is output from the shift register 206, and is output to 4 by the multiplier 212.
The signals are multiplied and added by the adder 215, and the adder 215 outputs the edge detect signal G · S11 as shown in the above-mentioned equation 2 and FIG.

Since the kernel elements on the n-2th line and the n + 2th line are the same, the shift registers 209 to 211 are omitted and n-
The outputs of the second line and the (n + 2) th line may be added and then input to the shift register 201. It is also clear that the edge detection unit 5 can be constructed by inserting the value of kernel * 4 into the multiplier using the circuit of FIG.

(Smoothing Processing Unit 3) FIG. 23 shows details of the blocks of the smoothing processing unit 3 shown in FIG.

The image signal S1 is continuous with 5 in the sub-scanning direction of the image.
It is composed of line data, and the sub-scan 5 is performed by the adder 271.
Pixel addition is performed. This data is input to the shift register 272 for delaying by 1 bit. The output data of the shift register 272 is input to the adders 277 to 280. The adder 272 adds the output of the shift register 272 and the data one pixel before. The result of this addition is 2
After being latched in the shift register 74, it is added to the next pixel by the adder 278. Similarly, at time T2, the adder 280 outputs S _N , m + as shown in FIG.
2 + S _N , m + 1 + S _N , m + S _N , m-1 + S _N , m
-2 is output (S _N , j = S _N −2, j + S _N −
1, j + S _N , j + S _N +1, j + S _N +2, j).

In this way, when the pixel of interest is Sn, m, the pixel sum indicated by * 3 in Expression 2 is output from the adder 280 and the divider 28
The smoothed data is obtained by dividing by the total number of pixels by 1. FIG. 26 shows a circuit for smoothing with weighting as shown in FIG. Operation timing is 23rd
Same as the figure, but in FIG. 26, multipliers 351 to 355 are shown.
Each line is weighted by the
.About.360 are weighted to perform smoothing as shown in FIG.

In this smoothing processing unit, all the added values are added in the image sub-scanning direction and then added in the image main scanning direction, so that the circuit scale can be reduced.

(Other Embodiments) In this embodiment, the kernels of the differential value detecting unit, the smoothing processing unit, and the edge enhancing unit are set to 5 × 5, but may be 3 × 3 depending on the number of lines for the purpose of removing moire. In some cases, 5 × 5 or more may be required. Depending on the purpose, the differential value detector,
The smoothing processing units do not need to use kernels of the same size. Furthermore, the kernel does not have to be square.

Further, in the present embodiment, a set of 5 line buffers is provided and the edge detection, smoothing and edge enhancement are performed by parallel processing, but it is not always necessary to perform them in parallel.

In this embodiment, the smoothing signal S6 output from the smoothing processing unit 3 and the edge enhancement signal S output from the adder 7 are output.
11 is added at a ratio according to the output of the gamma conversion unit 2, but the input image signal S1 is used instead of the edge emphasis signal S11.
May be used. In this case, although there are some inferiorities to the characters and line drawings as compared with the present embodiment, there is an advantage that the apparatus can be greatly simplified and the same effect as the present embodiment can be obtained with respect to suppression of moire.

Further, the edge enhancing section 5 and the multiplier 6 shown in FIG.
The edge enhancer configured by the adder 7 and the adder 7 does not require the multiplier 6 and the adder 7 when the center portion of the kernel * 4 of the edge enhancement unit 5 is configured to be variable by the constant S9.

Further, although the constant S9 is variable from the outside in the present embodiment, it may be fixed internally.

Further, in the embodiment of the present invention, the characteristic of the control signal generator 2 for performing the gamma conversion is explained as shown in FIG. 8. However, modified examples of the characteristic of the gamma converter 2 are shown in FIGS. Show.

Although only the characteristic of the control signal S4 is shown in FIG. 27, the control signal S3 is represented by S3 = 1−S4.

FIG. 27-a has the characteristics of S4 = 0, 0 <S2 <0.5, S4 = 1.0, 0.5 <S2 <1.0, and in particular, the circuit of the gamma conversion unit can be easily configured. There are features.

FIG. 27-b has a characteristic of S4 = -arctan (k · S2 + k), and is particularly characterized in that the connection between the smooth signal and the edge emphasis signal is smooth.

FIG. 27-c shows S4 = 0 where 0 <S2 <0.25 S4 = 0.33 where 0.25 <S2 <0.5 S4 = 0.67 where 0.5 <S2 <0.75 S4 = 1 However, it has a characteristic of 0.75 <S2 <1.0, which is relatively high compared to the embodiment shown in FIG.
The circuit is simple, and the connection between the smoothed signal and the edge emphasis signal is smoother than when the characteristics of the gamma conversion unit shown in FIG. 27-a are used.

Since it will not be further described, for example, Prewitt's edge detection or Sobel's edge detection method, which is well known as a differential value detection unit, may be used. Further, Laplacian may be used as the edge detection unit. Further, as the above-mentioned Prewitt's edge detection, Sobel's edge detection method or Laplacian, spatial filter processing is usually performed using a kernel of 3x3, but even if the size of the kernel is expanded to other than 3x3, it is essential to the present invention. It does not affect.

<Effect> As described above, according to the present invention, it is possible to prevent moire from occurring when a halftone image is reproduced, and at the same time, faithfully reproduce characters and thin lines by edge enhancement.

[Brief description of drawings]

FIG. 1 is a diagram showing a basic concept of an embodiment of the present invention, FIG. 2 is a frequency characteristic diagram of various images, FIGS. 3, 4, and 5 are frequency characteristic diagrams of various differential filters, and FIG. FIG. 7 is a frequency characteristic diagram of the smoothing filter, FIG. 7 is an image processing block diagram of the present embodiment, FIG. 8 is a gamma conversion characteristic diagram of the control signal generator 2, and FIGS. 9 and 10 are primary differentials. Differential value detector 1
FIG. 11 is a diagram showing an operation example of the smoothing processing unit 3, FIG. 12 is a diagram showing an operation example of the edge enhancing unit 5, and FIG. 13 is a signal waveform diagram of each unit in FIG. , FIG. 14 is a detailed circuit diagram of the edge detector 1, and FIG. 15 is a primary differentiator 30.
6, 312 are detailed block diagrams, FIG. 16 is a timing chart showing the operation of the primary differentiator, FIG. 17 is an absolute value circuit diagram, and FIG. 18 is a detailed block diagram of another example of the primary differentiator. FIG. 19 is a detailed block diagram of the edge enhancement unit 5, FIG.
FIG. 0 is a buffer circuit diagram, FIG. 21 is a diagram showing an image area, FIG. 22 is a diagram showing the operation of the edge enhancement unit, and FIG.
FIG. 24 is a detailed block diagram of the smoothing processing unit 3, FIG. 24 is a diagram showing the operation of the smoothing processing unit, FIG. 25 is a diagram showing a kernel for other smoothing processing, and FIG. 26 is FIG. 27 is a block diagram for performing the smoothing, and FIGS. 27 (a), (b), and (c) are diagrams showing other gamma conversion characteristics of the control signal generator 2. In the figure, a is an edge detector, b is an edge enhancer, c is a smoother, d
Is a mixer, 1 is a differential value detector, 2 is a control signal generator, 3
Is a smoothing processing unit, 4 and 6 and 8 are multipliers, 5 is an edge enhancement unit, and 7 and 9 are adders.

 ─────────────────────────────────────────────────── ─── Continuation of the front page (72) Inventor Yoshinobu Mita 3-30-2 Shimomaruko, Ota-ku, Tokyo Canon Inc. (56) Reference JP-A-57-24166 (JP, A)

Claims (1)

[Claims] 1. An edge detecting means for detecting an edge portion of an image signal, a smoothing means for smoothing the image signal, an edge emphasizing means for emphasizing an edge of the image signal, and the smoothing by the detection output of the edge detecting means. The conversion means or means for mixing or selecting the outputs of the emphasizing means and outputting the same,
An image processing apparatus characterized in that a peak frequency of a spatial filter constituting the edge detecting means is set to a lower frequency than a first harmonic frequency of a halftone image.