JPH07105359A - Picture processor - Google Patents

Abstract

PURPOSE:To provide the picture processor capable of an excellent resolution conversion at an optional magnification by eliminating jaggy having been already generated from the input without causing interpolation flurring and jaggy when information with low resolution is converted into information with high resolution. CONSTITUTION:Information with low resolution inputted from an input terminal 100 is interpolated by a linear interpolation means 101 and the interpolation information comprising (NXN) picture elements corresponding to a noted picture element is quantized by a quantization means 102. The quantization is equivalent to generation of edges and generated edge information and linear interpolation information are synthesized based on an adaptive ratio to avoid a picture of a picturesque tone by artificial edge generation while especially reducing the fog due to interpolation when the input is a natural picture. Furthermore, proper smoothing processing is executed before the interpolation to allow equivalent generation of edges with high resolution after edges of the original information are eliminated and excellent resolution conversion not generating jaggy in a character and a line picture or the like is attained.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an image processing apparatus, for example, an image output apparatus such as a printer for enlarging and scaling input image information and outputting the image information, and low resolution information to high resolution for communication between models of different resolution. The present invention relates to an image processing device and the like suitable for a device for converting resolution into information.

[0002]

2. Description of the Related Art Conventionally, various methods have been proposed as a method of converting input low resolution information into high resolution information. However, the proposed conventional method uses a target image type (for example, a multi-valued image having gradation information for each pixel, a binary image binarized by pseudo halftone, and a fixed threshold value of 2). The conversion processing method differs depending on the binarized binary image, character image, etc.).

An interpolation method which is a conventional resolution conversion method for a multi-valued image such as a natural image having gradation information for each pixel is the same pixel value as the observation point closest to the interpolation point as shown in FIG. Closest-point interpolation method of arranging 4 or 4 points surrounding the interpolation points as shown in FIG. 26 (pixel values of 4 points are A, B, C, D)
A co-linear interpolation method (linear interpolation method) for determining the pixel value E by the calculation of (Equation 1) according to the distance of is generally used.

[0004]

## EQU1 ## E = (1-i) (1-j) A + i. (1-j) B + j. (1-i) C + ijD (Equation 1) {However, when the pixel distance is "1" , A in the horizontal direction and j in the vertical direction from A. (I ≦ 1, j
≤1)}

[0005]

However, the above-mentioned conventional example has the following drawbacks. That is, the method of FIG. 25 has an advantage that the configuration is simple, but when the target image is used as a natural image or the like, the pixel value is determined for each block to be enlarged, and therefore the blocks are visually conspicuous. The image quality is poor.

Further, even when used for characters, line images, CG (computer graphic) images, etc., the same pixel value continues for each block to be enlarged.
As shown in FIGS. 27 (a) and 27 (b), the image becomes a jagged and conspicuous and inferior image. Although FIG. 27 shows an example of double resolution conversion in both the vertical and horizontal directions, the greater the magnification, the greater the deterioration (“2 in the figure”).
00 "and" 10 "are pixel values).

The method shown in FIG. 26 is a method which is generally well used for enlarging a natural image. According to this method, the image quality is averaged and smoothed, but the image quality is blurred in the edge portion and the portion where sharp image quality is required. Further, in the case of an image obtained by scanning a map or a natural image including a character portion, important information may not be transmitted to the receiver due to blurring due to interpolation.

FIG. 27 (c) shows the method of FIG.
7A shows image information obtained by interpolating the input image information of FIG. As is clear from FIG. 27C, the pixel values are not uniform not only around the diagonal lines but also around the diagonal lines, and blurring occurs. Furthermore, due to restrictions on application software on the host computer,
There is a case where jaggies have already occurred in the input low resolution information due to the processing such as the nearest-neighbor interpolation method (see FIG. 28).

As in this case, the means for eliminating the jaggies already generated in the input information could not be improved in any of the conventional examples.

[0010]

The present invention has been made for the purpose of solving the above-mentioned problems, and has the following structure as one means for solving the above-mentioned problems. That is, an image processing apparatus that converts low resolution information into high resolution information and increases the number of pixels of input image information by (N × M) times, in which one pixel of low resolution information is (N × M) pixels And a quantum for quantizing the image information in the block of (N × M) pixels after interpolation by the interpolation means corresponding to the pixel of interest on the low resolution into n values (n ≧ 2). And a calculating means for adding and synthesizing the information quantized by the quantizing means in units of (N × M) pixels after the interpolation by the interpolating means and the interpolation information from the interpolating means according to a set distribution ratio. Prepare

An image processing apparatus for converting low resolution information into high resolution information and increasing the number of pixels of input image information by (N × M) times, and smoothing means for smoothing the low resolution information. And interpolating means for interpolating one pixel of low resolution information into (N × M) pixels, and n values (n ≧ n) in the block of (N × M) pixels after interpolation corresponding to the target pixel on the low resolution.
2) and a quantizing means for quantizing.

[0012]

With the above construction, when converting the input low resolution information into the high resolution information, blurring due to interpolation, characters, whole images, etc., which have been particularly problematic for natural images, have been particularly problematic. It is possible to realize a conversion process with good image quality without causing jaggies. Further, due to the limitation of the application software that created the image, the jaggy-filled curve can be converted into a smooth curve even when an image in which jaggies have already occurred is input.

It is also possible to create an edge in the interpolation information by quantizing, and to change the nonlinearity easily by adaptively changing the composition ratio of the created edge information. In particular, when the input is a natural image, it is possible to avoid becoming a painterly image due to artificial edge creation.

[0014]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS An embodiment according to the present invention will be described in detail below with reference to the drawings. (First Embodiment) FIG. 1 is a block diagram showing a main part of a first embodiment according to the present invention. The image processing apparatus of this embodiment is
Although it is efficient to equip an image output device such as a printer mainly with an image processing device other than the image output device,
It is also possible to incorporate it as application software in the host computer.

The operation procedure of this embodiment will be described with reference to the block diagram of FIG. In the present embodiment, an example will be described in which input image information is converted into information of the number of pixels N times vertically and M times horizontally (N and M are each an integer of 2 or more). 100 in the figure
Indicates an input terminal to which low-resolution image information is input. This low resolution information is transmitted to the linear interpolation means 101, and
By the next interpolation process (hereinafter referred to as "linear interpolation process"),
The pixels between the original samplings are filled in, and interpolation information of vertical N times and horizontal M times is created. FIG. 26 shows the linear interpolation processing.
It is the same as the processing shown in. Reference numeral 102 denotes a quantizer, which is an n-value (n-value) block for each block of interpolation information of N pixels × M pixels centered on a pixel of interest (referred to as Ixy) of low resolution information.
≧ 2) shows the means for quantization. This is equivalent to creating an edge for each low resolution pixel by quantizing the interpolation information. Details of this edge creation will be described later (hereinafter, the information after quantization is referred to as “edge information”).

Reference numeral 103 in the figure denotes a distribution ratio determining means. This is a means for calculating the distribution ratio (a, where 0 ≦ a ≦ 1) in combining the created linear interpolation information and edge information. This distribution ratio is also determined for each block of N pixels × M pixels. The details of the determination of the distribution ratio will be described later. After that, the edge information is multiplied by a in the multiplier 104 and the linear interpolation information is multiplied by (1-a) in the multiplier 105 using the obtained distribution ratio a, and then combined in the adder 106. Reference numeral 107 in the figure denotes an output terminal, and the input image information is converted into N × M times information and output.

FIG. 2 is a diagram showing a detailed configuration of the quantizing means 102 shown in FIG. In the present embodiment, description will be made by taking binarization (n = 2) for quantization as an example. The part surrounded by the broken line corresponds to the quantizing means. Reference numeral 110 in the drawing denotes an input terminal for low resolution information, to which the information indicated by 111 is input. A to I
Are low resolution information. In 111, the pixel E corresponds to the target pixel, and the portion surrounded by the alternate long and short dash line is the window for the neighboring pixels of the target pixel (reference window and the size of the reference area around the target pixel). 11
Reference numeral 2 denotes an input terminal from the linear interpolation means, to which the information shown in FIG. 3 is input. Here, an example of N = M = 5 is shown.

In FIG. 3, a broken line indicates a block boundary centered on each low-resolution pixel, and a portion surrounded by a solid line is a block for the target pixel E (hereinafter referred to as "target pixel block"). Further, ◯ marks indicate pixels (observation pixels) of low resolution information, and X marks indicate interpolation pixels. That is, it is considered that the low resolution pixel E corresponds to 25 × 5 × 5 pixels by interpolation.

With respect to the low resolution information input from the input terminal 110, the maximum and minimum values in the window are detected by the MAX / MIN detecting means 113. MAX information detected, M
The IN information is transmitted to the threshold value determining unit 114, and the threshold value that is quantized into binary values is determined. In this embodiment, the threshold value (“T
H ”) is determined by the following formula. TH = (MAX + MIN) / 2 (Equation 3) The determined threshold information, MAX information, and MIN information are transmitted to the binarizing means 115, and the pixel E of interest subjected to the linear interpolation, which is input from the input terminal 112, is input. The interpolation information (within the pixel block of interest) of the center block is binarized. The MAX value is assigned to the interpolation pixel larger than the binarization threshold, and the MIN value is assigned to the small interpolation pixel. Reference numeral 116 denotes an output terminal, and the binary representative value as shown in 117 is MA.
The edge information of the blocks assigned by X and MIN is output.

FIG. 4 shows a state of linear interpolation and edge creation. In order to simplify the explanation, it is shown in a one-dimensional direction. The ∘ mark indicates the pixel value of the sampling point on the low resolution, and the X mark indicates the pixel value of the interpolation point interpolating between them.
The circle marked in the center is the pixel of interest. In the figure,
As shown in (a), the values of MAX and MIN are detected from the adjacent pixels, and the threshold value TH is calculated by the above-mentioned formula. Now, assuming that the enlarged pixel block centered on the pixel of interest is 5 pixels, as shown in FIG.
Since it is H or more, it takes the value of MAX, and since one pixel is less than TH, it takes the value of MIN.

FIG. 5 shows an example of creating edge information. Now, it is assumed that the pixel of interest of the low resolution information has a pixel value of "80" arranged in the center as shown in FIG. One pixel of this pixel of interest is interpolated vertically N times and horizontally M times as shown in FIG. 5B, and MAX value “200”, MIN value “20”, and binarization threshold value ( 200+
20) / 2 = 110 and binarized to create an edge.

By the edge creating process described above, it is possible to create a high-resolution information, that is, a pixel in which a pixel caught in an edge portion in a low-resolution image is smooth in the resolution direction in the enlarged block. become. Figure 6
Indicates a distribution ratio determining means. In the figure, the part surrounded by the broken line corresponds to the distribution ratio determining means indicated by 103 in FIG. In the figure, reference numeral 121 denotes an input terminal for inputting the low resolution information shown at 122. As described with reference to FIG. 2, the pixel E is set as the pixel of interest. Similar to the edge creating means in FIG. 2, the MAX / MIN detecting means 123 detects the MAX value and the MIN value in the window. Of course, this detecting means can be shared with that in the edge creating means.

Reference numeral 124 denotes a subtractor, which is (MAX-MI
N) is calculated. That is, it corresponds to obtaining the dynamic range in this window. 125
Is a weighting means, and is a means provided for deciding whether the created edge information should be emphasized more or less. This weighting may be experimentally obtained in order to optimize the system, or may be determined according to the target image.

FIG. 7 shows an example of weighting. Fig.7 (a)-
All of (d) shows how to calculate the distribution ratio a based on the dynamic range in the window. As the dynamic range increases, the weight (coefficient a) of the edge creation information increases, but it is linear (see FIG. 7).
(A)), non-linear one (Fig. 7 (b)), clipped one (Fig. 7 (c)), one set on stairs (Fig. 7).
Various weightings such as (d)) are possible. Also, combinations of these are possible. These may be obtained by calculation, or a complicated one may be realized by an LUT (lookup table). The distribution ratio a thus obtained is output from the output terminal 126, and manages the composite distribution of the edge information and the linear interpolation information.

FIG. 8 is a diagram showing the effect of combining the edge information by the distribution ratio and the linear interpolation information. That is, the created edge information is multiplied by a, and the linear interpolation information is multiplied by (1-a) and combined to create a smooth edge as shown in the figure. Moreover, as described above, since the dynamic range, that is, the degree of the edge size is used as the evaluation function, it is possible to perform processing in which the created edge information strongly depends on the larger edge ((a in FIG. 8). )reference). Further, since the linear interpolation information becomes more dependent on the flat part, the created edge does not have a bad influence (see FIG. 8B).

As described above, according to the present embodiment, when the input low resolution information is converted into high resolution information, blurring due to interpolation, which has been a particular problem in natural images, etc.,
It is possible to realize a conversion process with good image quality without causing jaggies, which has been a particular problem in characters and all images. Further, due to the limitation of the application software that created the image, the jaggy-filled curve can be converted into a smooth curve even when an image in which jaggies have already occurred is input.

Further, it corresponds to easily changing the non-linearity by adaptively changing the composition ratio of the created edge information, and particularly when the input is a natural image,
It is possible to avoid a painting-like image due to artificial edge creation. (Second Embodiment) Next, a second embodiment according to the present invention will be described in detail. The second embodiment differs from the above-described first embodiment in the distribution ratio determining means, that is, the evaluation function for determining the distribution ratio. FIG. 9 is a principal block diagram showing a second embodiment according to the present invention, showing a distribution ratio determining means in the second embodiment. In FIG. 9, FIG.
Similarly, the part surrounded by the broken line shows the distribution ratio determining means. 6, those parts that are the same as those corresponding parts in FIG. 6 are designated by the same reference numerals, and a description thereof will be omitted.

In FIG. 9, reference numeral 130 denotes an input terminal,
The low resolution information shown in 131 is input. In 131, the pixel of M is set as the pixel of interest, the eight adjacent pixels surrounded by the bold line are referred to as eight neighborhoods, and the peripheral pixels surrounded by the one-dot chain line are two.
4 neighborhoods. 132 is MAX near 8
MIN detection means, 133 is MAX / MIN near 24
As the detection means. Subtractors 134 and 135 calculate the difference between the detected MAX and MIN, respectively. This process includes the dynamic range in 8 neighborhoods and 2 including it.
This is equivalent to obtaining the dynamic range in the vicinity of 4. Based on the respective obtained dynamic ranges, the edge angle determination means 136 determines how steep the edge is. FIG. 10 shows how the judgment is made.

In FIG. 10, the ∘ mark indicates the pixel value of the sampling point on the low resolution, and the X mark is the interpolation value for interpolating between the pixels. Now, in order to simplify the explanation, the explanation will be given in the one-dimensional direction. The circle marked in the center is the pixel of interest,
Pixels on both sides adjacent to each other are 8 neighbors, and pixels on the outside are 2
The number of pixels is four. 10A shows a case where the difference between the dynamic range near 8 and the dynamic range near 24 is small, and FIG. 10B shows a case where the difference is large.

Based on these two types of dynamic range information, the rough angle of the edge surrounding the pixel of interest is determined,
A signal is transmitted to the distribution ratio creating means 137. This may be the difference between the two types of dynamic ranges, or the ratio is effective because the relative relationship is evaluated. Allocation ratio creating means 1
In 37, when the edge angle is large, the distribution ratio of the created edge information is increased, and when the edge angle is small, the distribution ratio of the created edge information is decreased. decide. The determined distribution ratio is output to the output terminal 126, and the signals are given to the multipliers 104 and 105 as in the first embodiment shown in FIG.

As described above, according to the second embodiment, the non-linearity can be easily changed by adaptively changing the composition ratio of the created edge information as in the first embodiment. In particular, when the input is a natural image, it is possible to avoid a painterly image due to artificial edge creation. (Third Embodiment) FIG. 11 is a principal block diagram showing a third embodiment according to the present invention. In the third embodiment, the window creating means 1 is added to the configuration of the first embodiment shown in FIG. 1 described above.
40 is added. That is, MAX and MIN in the edge creation in the quantizing means 102.
The window used for the calculation of is variably configured. As described in the above embodiment, it may not be said that the edge peaks are detected only by the windows in the vicinity of 8.

Therefore, in the third embodiment, the window creating means 140 adaptively creates a window including the edge peaks according to the surrounding pixel value, and M in the window created by the window creating means 140 is adjusted.
AX · MIN is used to create the threshold TH. As a result, a window including the edge peaks can be adaptively created, and the edge peaks can be reliably detected.

It is also possible to supply MAX / MIN in the window created by the window creating means 140 to the allocation ratio determining means 103 and use it for determining the allocation ratio. In the first to third embodiments described above, the size of the window for calculating the threshold value and the distribution ratio may be set in advance, such as fixed in the vicinity of 8 or fixed in the vicinity of 24, depending on the type of the input image. good.

Further, the distribution ratio may be changed according to the type of input image and the preference of the user. Further, although the binarization by the threshold value described above has been described as the creation of the edge, it goes without saying that another method may be used. In addition, although the binarization (n = 2) has been described for the quantization, similarly, the quantization representative value and the threshold value are set from the window of the peripheral pixels, and two or more values are set in the pixel block of interest after interpolation processing. It may be converted into a multi-value of. It is also possible to reduce the steepness of the edge by a multi-valued operation.

Further, in order to facilitate the explanation, 2
Although the creation of the binarized edge image, the creation of the linear interpolation information, and the synthesis of the two kinds of images have been described independently, in the actual processing, the information after the linear interpolation causes the pixels above the threshold value TH to be Levelout. = A * MAX + (1-a) * Level
For pixels less than line TH, Levelout = a × MIN + (1-a) × Level
lin {however, Levelin is each pixel value after linear interpolation, Le
Velou is the output value at the same pixel position, and a is the distribution ratio (0 ≦
a ≦ 1). } Of course, it can be calculated at the same time.

As described above, according to the first to third embodiments, the edge information is speculatively created from the input low resolution information and the information is added, so that there is no visually blurred feeling. You can create high resolution information with sharp edges. It also corresponds to easily changing the non-linearity by adaptively changing the composition ratio of the created edge information. Especially when the input is a natural image, a painting-like image created by artificial edge creation. Can be avoided.

(Fourth Embodiment) FIG. 12 is a principal block diagram showing a fourth embodiment according to the present invention. The first to third embodiments described above are mainly effective for resolution conversion of natural images, but the fourth embodiment mainly uses PDL (page description language) on the host computer and various It is effective for characters, line images, etc. created by application software.

The operation procedure of the fourth embodiment will be described with reference to the block diagram of FIG. In the figure, reference numeral 100 denotes an input terminal to which low resolution image information is input. The input image information is transmitted to the smoothing means 150. FIG. 13 shows an example of the smoothing filter used in the fourth embodiment. Of course, other smoothing filters may be used. The image information smoothed by the smoothing means 150 is the linear interpolation means 101.
Then, the smoothed image information is subjected to interpolation of N times in the vertical direction and M times in the horizontal direction.

The information after the interpolation is transmitted to the quantizing means 102, and an edge is created in the pixel block of interest in which the number of pixels has increased, as in the above-described embodiment. This operation destroys the edge of the original information (low resolution information) and creates a new edge (high resolution information) with the number of pixels increased, so an edge where jaggies that depend on the original information occurs Absent.

The processing of the fourth embodiment will be described with reference to FIG. 14 based on actual values. FIG. 14A shows a part of a certain input image. Now, the pixel surrounded by the broken line is the target pixel, and the part surrounded by the alternate long and short dash line is the window. 14
(B) indicates the pixel value of the same portion as (a) after smoothing. Since the pixel of interest indicated by the broken line is smoothed, “200” is converted to a value of “150”.

FIG. 14C shows pixel values after linear interpolation near the target pixel. Now, with N = M = 3, one pixel is increased to 9 pixels. In FIG. 14C, 9 pixels surrounded by a broken line are the target pixel block. The inside of this pixel block of interest is quantized into a binary value. 14
From the window of (a), MAX = 200, MIN = 5
Since it is 0, the binarization threshold value is binarized with TH = 125, and the MAX value is substituted for the pixels above TH and the MIN value is substituted for the pixels below TH. The result is shown in FIG. FIG. 14E shows the final result after performing the same processing on the other target pixels. As is clear from this example, good resolution conversion can be realized at any magnification without causing jaggies.

(Fifth Embodiment) FIG. 15 is a block diagram showing the essential parts of a fifth embodiment of the present invention. In the figure, the same parts as those in the above-mentioned embodiment are designated by the same reference numerals for description. 1 in the figure
Reference numeral 60 denotes a filter selection means, which is a means for adaptively selecting which filter is used as a smoothing filter from the input image. For example, if the smoothing filter shown in FIG. 13 is used for the filtering process, the edges are grouped together at a portion where a plurality of edges are mixed, or the line is broken or disappears at a thin line portion. However, in the fifth embodiment, the filters are switched according to the pattern in the window of the input image in order to solve the problem.

In FIG. 15, filter selecting means 160 is provided before the smoothing means 150, and a filter group 161 is provided.
Determine which filter to select. FIG. 16 shows an example of the filter selection means 160. In FIG. 16, reference numeral 170 denotes an input unit, to which the pixel information in the window around the pixel of interest shown at 171 is input. Reference numeral 172 denotes MAX / MIN detecting means, which detects the maximum and minimum values in the window. Reference numeral 173 denotes a threshold value determining means,
The threshold value TH is calculated from the values of MAX and MIN. Reference numeral 174 denotes a binarizing unit that binarizes not only the pixel of interest but also each pixel in the window. For example, in the window of 171, nine pixels (A to I) in the window are binarized. The binarized 9-bit window information is transmitted to an LUT (look-up table) 175 and an optimum filter is selected (176 is an output terminal).

In the LUT 175, a pattern that causes a harmful effect due to smoothing is experimentally obtained in advance, and a filter weighted to the pixel of interest as shown in FIG. 17 is set or the filter is set to be through. Good. Figure 1
FIG. 8 shows a typical pattern that may cause harmful effects in the filter of FIG. 18A and 18B are fine line patterns, and there is a possibility that the fine lines may be cut due to smoothing. 18C and 18D are patterns corresponding to the corners, and there is a possibility that the corners may be rounded by the smoothing. In the fifth embodiment, by setting such various patterns in the LUT, an adaptive filter can be selected, the above-mentioned harmful effects can be prevented, and jaggies may occur. It is possible to process only.

The MAX / MIN detecting means shown in FIG.
It goes without saying that the threshold value determining means, the binarizing means, etc. can be shared with the quantizing means (configuration of FIG. 2) shown in FIG. It is efficient to store Further, in the fifth embodiment, the filter switching is realized by the binary pattern in the window, but it is also possible to control the filter switching by calculation by using the complexity of the edge around the target pixel as an evaluation function. .

(Sixth Embodiment) FIG. 19 is a principal block diagram showing a sixth embodiment according to the present invention. In this example,
At the time of the input low resolution information, it is particularly effective for image information that already has jaggies due to the limitation of the image input device, the limitation of the application software that created the image, and the like. The switch 180 is initially connected to the terminal A, and the input image information is transmitted to the smoothing means 150. After the smoothed image information is transmitted to the linear interpolation means 101, the interpolation information is quantized by the quantization means 102, and an edge is created in the expanded information.

The created edge information is connected to the terminal B in the switch 181, and the linear thinning means 18
2 is sent. FIG. 20 shows a method of linear thinning processing. For simplification of description, FIG. 20 is shown in a one-dimensional direction. ○ indicates the pixel of high resolution information that created the edge,
The crosses indicate sampling pixels for linear thinning. Although FIG. 20 shows an example of decimating to 1/2, this decimating process is executed until the resolution is the same as the resolution of the image input through the input terminal 100 of FIG.

When the thinning process is performed and the resolution becomes the same as that of the input image, the switch 180 is connected to the terminal B, and the same operation is executed again. Switches 180, 181
Is controlled by the path counter 183, and a loop of smoothing, edge creation, and thinning processing is executed for a preset number of paths. If you repeat this process for the set number of passes,
The enlarged image information is output to the output terminal 107 by being connected to the terminal A by the switch 181. By this series of iterative processing, the jaggy that originally existed is smoothed and a new edge is created repeatedly, so
Jaggies can be transformed into smooth curves. Moreover, since an edge is created, the image is not blurred by the interpolation process.

FIG. 21 shows an example of the processing result according to the sixth embodiment. This is assumed, for example, when a line image created by application software on a host computer is input. For ease of explanation, the unit cell surrounded by a square is one pixel. Jagging has already occurred from the input low resolution information {(a) in FIG. 21}.

FIG. 21B shows the processing result of the iterative processing of smoothing and edge creation of the sixth embodiment. The unit lattice is smaller than that of (a) because it is converted to high resolution. As described above, according to the sixth embodiment, iterative processing can improve the jaggies generated from the original information into smooth curves.

(Seventh Embodiment) FIG. 22 is a principal block diagram showing a seventh embodiment according to the present invention. In the figure, the same parts as those in the above-mentioned embodiment are designated by the same reference numerals for description. In the seventh embodiment, compared with the sixth embodiment shown in FIG.
The control of the number of loop iterations is different. In FIG. 22, reference numeral 190 denotes a repetitive determination means, which is an input terminal 1
Two kinds of image information, that is, the original image information input by 00 and the interpolated image information in which an edge is created and enlarged are input.

The repetition determining means 190 compares the image information of so-called pre-value interpolation, in which the pixel values of the original image information are simply repeatedly arranged by the enlargement magnification, with the image information of which the edge has been created, and compares them sufficiently. The switching of the switches 180 and 181 is controlled by determining whether the jaggy has been cut off. That is, while the jaggies cannot be removed, the switches 180 and 181 are connected to the terminal B and are repeatedly processed.

If it is determined that the jaggy has been cut off, the switch 181 is connected to the terminal A, and the processed image is output to the output terminal 107. As an example of the determination, whether or not the error power between the above-described previous value interpolation image information and the edge creation information is equal to or more than a certain set threshold value may be used as the evaluation function of the determination. As described above, according to the seventh embodiment, as in the sixth embodiment described above, by repeating smoothing, enlargement, edge creation, and thinning, jaggies are reduced, and instead a smooth curve is obtained. Formed, and finally a good image is output.

(Eighth Embodiment) FIG. 23 is a principal block diagram showing an eighth embodiment according to the present invention. In the figure, the same parts as those of the sixth embodiment shown in FIG. In the eighth embodiment, the quantizing means 200 is partially different from the sixth embodiment. FIG. 24 shows the outline of the quantizing means of the sixth embodiment. This is different from the quantizing means shown in FIG. 2 only in that the information of MAX and MIN, which are the quantized representative values in the binarizing means 210, is inputted from the outside, and the others are the same. There is (the broken line portion is the quantizing means 200). That is, in the embodiment shown in FIG. 19, the density of the original image is averaged by the linear decimation means, the dynamic range becomes smaller, and the difference in density from the original image becomes larger and larger due to the repeated processing. There is a risk of becoming. On the other hand, in the eighth embodiment,
It is a means for improving the above-mentioned problems, and MAX and MIN for determining the threshold value for creating an edge use the information by the window after linear thinning (the first time is the original information),
MAX and MI which are the binarized representative values after the edges are created
N is characterized by using the information of the original image (input terminal 211 in FIG. 24). By this processing, the binarized quantized representative value is always the pixel value of the original image information, and the side effect of the density change can be avoided only by improving the jaggies in the resolution direction.

In the eighth embodiment, the MAX and MIN values are input from the original image information every time during the iterative process, but the present invention is not limited to the above example, and the iterative process can be performed. In the case of the last loop, for example, N loops, MAX and MIN of the original image are used only in the Nth processing, and (N−
1) In the processing up to the first time, it is also effective to use MAX and MIN after linear thinning (the first time is original information) as in the embodiment of FIG.

In the above description of the embodiments, the interpolation means has been described by taking linear interpolation as an example, but the invention is not limited to this. In the fourth to eighth embodiments described above, the edge creating means is {(MAX + MI
Although binarization by the threshold value of (N) / 2} has been described, it goes without saying that other edge creating means may be used.

Also, an embodiment has been described in which the smoothing filter is adaptively switched based on the partial characteristics of the input image.
It is also effective to switch the filter according to the number of times of repeated processing. As described above, according to each of the fourth to eighth embodiments, jaggies that occur when converting input low resolution information into high resolution information, restrictions on image input devices, application for creating images, etc. It is possible to convert jaggies that have already occurred due to software restrictions to smooth curves.

Further, it is possible to limit smoothness by controlling the number of repetitions. Further, by adaptive filter processing, it is possible to distinguish the portion to be converted into a curve. As described above,
According to each of the eighth embodiments, since low-resolution image information can be easily converted into high-resolution information, inter-model communication with different resolutions, a printer that performs enlargement / magnification, and outputs a high-quality image, By applying it to a copying machine, it becomes possible to output high-quality images to each.

The present invention may be applied to a system composed of a plurality of devices or an apparatus composed of a single device. Further, it goes without saying that the present invention can be applied to the case where it is achieved by supplying a program to a system or an apparatus.

[0060]

As described above, according to the present invention, jaggies that occur when input low resolution information is converted to high resolution information, restrictions on image input devices, restrictions on application software that creates images, etc. The jaggies that have already occurred can be converted into smooth curves by the above. Also, by controlling the number of repetitions,
It is possible to control the smoothness, and it is possible to divide the portion to be converted into a curve by adaptive filter processing.

Furthermore, according to the present invention, edge information is estimated and created from the input low resolution information, and the information is added, so that high-resolution information with sharp edges that does not visually blur is created. can do. It also corresponds to easily changing the non-linearity by adaptively changing the composition ratio of the created edge information. Especially when the input is a natural image, a painting-like image created by artificial edge creation. Can be avoided.

As described above, according to the present invention, the low resolution image information can be easily converted into the high resolution information, so that the present invention can be applied to various devices. For example, communication between models having different resolutions and enlargement conversion. It is possible to provide a printer or a copying machine that doubles and outputs a high-quality image.

[Brief description of drawings]

FIG. 1 is a block diagram of main parts showing a first embodiment according to the present invention.

FIG. 2 is a block diagram of a main part showing a quantization unit of FIG.

FIG. 3 is an explanatory diagram of linear interpolation according to the present embodiment.

FIG. 4 is a diagram for explaining linear interpolation and binarization according to the present embodiment.

FIG. 5 is an example of binarization according to the present exemplary embodiment.

FIG. 6 is a diagram showing a detailed configuration of a distribution ratio determining unit in FIG.

FIG. 7 is a diagram illustrating an example of the weighting means in FIG.

FIG. 8 is a diagram illustrating a difference in distribution ratio depending on the size of an edge.

FIG. 9 is a diagram of a distribution ratio determining unit according to a second embodiment of the present invention.

FIG. 10 is an explanatory diagram of a difference in edge angle according to the second embodiment.

FIG. 11 is a principal block diagram showing a third embodiment according to the present invention.

FIG. 12 is a principal block diagram showing a fourth embodiment according to the present invention.

FIG. 13 is an example of a smoothing filter according to a fourth embodiment.

FIG. 14 is a processing example in the fourth embodiment.

FIG. 15 is a principal block diagram showing a fifth embodiment according to the present invention.

16 is a detailed configuration diagram of a filter selection unit in FIG.

17 is an example of a smoothing filter in FIG.

18 is an example of a window pattern after binarization in FIG.

FIG. 19 is a principal block diagram showing a sixth embodiment according to the present invention.

FIG. 20 is an explanatory diagram of linear thinning processing according to the sixth embodiment.

FIG. 21 is a diagram for explaining the effect of the sixth embodiment.

FIG. 22 is a principal block diagram showing a seventh embodiment according to the present invention.

FIG. 23 is a principal block diagram showing an eighth embodiment according to the present invention.

FIG. 24 is a block diagram of the essential parts of the quantizing means of the eighth embodiment.

FIG. 25 is a diagram showing a nearest neighbor interpolation method as a conventional example.

FIG. 26 is a diagram showing a conventional co-linear interpolation method.

FIG. 27 is a diagram showing a processing example of a conventional example.

FIG. 28 is a diagram showing an example of input image information.

[Explanation of symbols]

 100 input terminal (low resolution image input) 101 linear interpolation means 102 quantizing means 103 distribution ratio (a) determining means 104, 105 multiplier 106 adder 107 output terminal (N × M times image information) 110 input terminal 111 low resolution Information 112 Input terminal 113 MAX MIN detection means 114 Threshold value determination means 115 Binarization means 116 Output of edge creation means 117 Edge information 121 Input terminal 122 Low resolution information 123 MAX MIN detection means 124 Subtractor 125 Weighting means 126 Output terminal 130 Input terminal 131 Low resolution information 132 8 Neighborhood MAX MIN detection means 133 24 Neighborhood MAX MIN detection means 134, 135 Subtractor 136 Edge angle determination means 140 Window creation means 150 Smoothing means 160 Filter selection means 160 Fill Group 170 input terminal 171 low resolution information 172 MAX MIN detection means 173 threshold value determination means 174 binarization means 175 LUT (look-up table) 176 output terminals 180, 180 switches 182 linear thinning means 183 path counter 190 cut-back determination means 200 Quantizer 210 Binarizer 211 Input terminal

Claims (17)

[Claims] 1. A method for converting low resolution information into high resolution information,
An image processing device for increasing the number of pixels of input image information by (N × M) times, and an interpolation means for interpolating one pixel of low resolution information into (N × M) pixels, and attention to low resolution. The image information in the block of (N × M) pixels after the interpolation by the interpolation means corresponding to the pixel is set to n value (n
≧ 2), and a distribution ratio in which the quantizing means for quantizing is set, and the information quantized by the quantizing means in units of (N × M) pixels after being interpolated by the interpolating means and the interpolating information from the interpolating means are set. An image processing apparatus comprising: 2. The image processing apparatus according to claim 1, wherein the quantized representative value after being quantized by the quantizing means in the computing means is calculated from the peripheral pixels of the low resolution pixel of interest. 3. The image processing apparatus according to claim 2, wherein when the quantization is binarization, the maximum value and the minimum value in the peripheral pixel group are assigned to the quantization representative value of the binarization. 4. The image processing device according to claim 1, wherein a threshold value for quantization in the quantization means is calculated from peripheral pixels of a low resolution pixel of interest. 5. The image processing apparatus according to claim 4, wherein when the quantization is binarization, a binarization threshold is an average of the maximum value and the minimum value. 6. The image processing apparatus according to claim 1, wherein the distribution ratio in the calculation means is determined by a pixel value width around a target pixel on a low resolution. 7. The image processing apparatus according to claim 1, wherein the distribution ratio in the calculation means is determined by a change in a pixel value width around a target pixel on a low resolution. 8. The image processing according to claim 1, wherein the size of the reference area around the target pixel on the low resolution that determines the distribution ratio in the calculation means is variable. apparatus. 9. The size of a reference area around a pixel of interest on a low resolution that determines the quantized representative value in the calculating means is variable. Image processing device. 10. The size of a reference area around a target pixel on a low resolution that determines a quantization condition such as a threshold value in the quantization means is variable. The image processing device according to item 1. 11. An image processing apparatus for converting low resolution information into high resolution information and increasing the number of pixels of input image information by (N × M) times, the smoothing means smoothing the low resolution information. And an interpolating means for interpolating one pixel of low resolution information into (N × M) pixels, and n values (n ≧ n) in the block of (N × M) pixels after interpolation corresponding to the target pixel on the low resolution. 2) An image processing device having a quantizing means for quantizing. 12. A thinning means for thinning out the high-resolution information after quantization is further provided, and each processing of the smoothing means, the interpolating means, the quantizing means, and the thinning means is repeatedly executed a plurality of times. The image processing apparatus according to claim 11, characterized in that. 13. The image processing apparatus according to claim 12, wherein the number of repetitions is controlled by comparing the image after the repeated processing with the unprocessed image. 14. The image processing apparatus according to claim 11, wherein the smoothing processing of the smoothing unit is adaptively switched according to a pixel value around a target pixel on a low resolution. . 15. The image processing apparatus according to claim 12, wherein the smoothing process of the smoothing unit is switched according to the number of times of repetition. 16. A pixel value around the pixel of interest is an n-value (n
16. A smoothing condition of the smoothing means is switched depending on a pattern after the quantization by the quantizing means. The image processing device according to item 1. 17. The method according to claim 11, wherein the interpolation means performs interpolation processing by linear interpolation.
The image processing device according to any one of 1.