JPH11331565A - Device and method for processing image - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide a device and a method for processing image with which the inputted image information of a low resolution is converted to the high- definition image information of a high resolution. SOLUTION: A maximum/minimum value detecting part 103 detects plural representative values from the pixel values of peripheral pixels including the concerned pixel of image information inputted from an input terminal 100 to a line buffer 101 and made into window by a window part 102, an interpolation function operating part 104 finds the interpolation values of plural interpolation pixels to interpolate the concerned pixel, a product sum coefficient calculating part 105 calculates respective product sum coefficients corresponding to the contrast of plural representative values, and a product sum operating part 106 operates the output interpolation values of the plural interpolation pixels through the product sum operation of respective product sum coefficients, plural representative values and interpolation values of plural interpolation pixels and outputs them from an output terminal 107.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an image output device such as a printer for enlarging and changing input image information and outputting the image information, and a resolution conversion from low-resolution information to high-resolution information by communication between models having different resolutions. The present invention relates to an image processing apparatus and method.

[0002]

2. Description of the Related Art Conventionally, various methods have been proposed for converting the resolution of inputted low-resolution information into high-resolution information. In the proposed conventional method, the type of the target image (for example, a multi-valued image having gradation information for each pixel, a binary image binarized by pseudo halftone, a binarized image by a fixed threshold) Binary image, character image, etc.)
The conversion processing method is different.

As a conventional interpolation method, a closest interpolation method in which the same pixel values closest to the interpolation point are arranged as shown in FIG. 11 and four points (4) surrounding the interpolation point as shown in FIG. A bilinear interpolation method is generally used in which the pixel value E is determined by the following calculation based on the distance between the pixel values of points (A, B, C, and D).

E = (1−i) (1−j) A + i · (1−j) B + j · (1−i) C + ijD Equation 1 (However, when the distance between pixels is “1”, It is assumed that there is a distance i in the horizontal direction and j in the vertical direction (i ≦ 1, j ≦
1)).

Further, as represented by the sampling theorem for a long time, a conversion processing method for converting a sampled discrete signal by a continuous signal is reproduced by passing through an ideal low-pass filter that can be expressed by a SINC function. There is a way to do that. Since it takes a long processing time to calculate the SINC function, there is a method of approximating an interpolation function represented by the SINC function and calculating an interpolation value only by a simple product-sum operation.

According to "Image Analysis Handbook: Mikio Takagi and Hirohisa Shimoda, University of Tokyo Press", an approximation of an interpolation function can be realized in the third-order convolution interpolation method (Cubic Convolution interpolation). Using the image data of 16 observation points around the point to be interpolated, the image data to be obtained is interpolated using a cubic convolution function represented by the following equation.

[0007]

(Equation 1)

[[] Is a Gaussian symbol and takes an integer part) P
11 to P44 indicate peripheral pixel values, and the arrangement is shown in FIG.

[0009]

However, the above-mentioned prior art has the following drawbacks.

First, although the nearest neighbor interpolation method has an advantage that its configuration is simple, when a target image is used for a natural image or the like, a pixel value is determined for each block to be enlarged, so that it is visually determined. The blocks are conspicuous and the image quality is poor.

Even when used for characters, line images, CG (computer graphic) images, etc., the same pixel value continues for each block to be enlarged.
The image becomes jaggy and jagged. FIGS. 14 and 15 show the appearance of jaggies.
Shown in FIG. 14 shows an example of input information, and FIG. 15 shows an example of resolution conversion in which the number of pixels is doubled both vertically and horizontally by the nearest neighbor interpolation method. Generally, the higher the magnification, the greater the image quality degradation ("200" and "10" in the figure are pixel values).

[0012] The bilinear interpolation method is a commonly used method for enlarging a natural image. In this method, image quality is averaged and smoothed, but blurred image quality occurs at an edge portion or a part where sharp image quality is required. Further, in the case of an image obtained by scanning a map or the like or a natural image including a character portion, important information may not be transmitted to a recipient due to blurring due to interpolation.

FIG. 16 is a diagram showing image information obtained by interpolating the input information shown in FIG. 14 twice vertically and horizontally by bilinear interpolation. As is clear from FIG. 16, the pixel values are not uniform not only in the vicinity of the oblique line but also in the oblique line itself.
Blur will occur.

In the third-order convolution interpolation method approximating the interpolation function, generally good resolution conversion and enlargement processing can be realized.
However, when used in the enlargement processing, if the enlargement ratio becomes large, blurring due to interpolation naturally occurs although the smoothing is not performed as in the bilinear interpolation method.

FIG. 17 shows the state of interpolation as viewed from the frequency domain. It is shown in one dimension for ease of explanation.

FIG. 17A shows a band of a signal sampled for each spatial distance S. Naturally, since the signal is a discrete signal, the signal is repeated every 1 / S. This is assumed to be an input signal form sampled every S. FIG. 17B shows an ideal low-pass filter by a broken line. This corresponds to the interpolation function. Ideally -1 / 2S
Since only the band up to １ ／ S is passed, the signal after filtering is as shown in (c) shown in FIG.

This is the restoration of a continuous signal using an interpolation function. (D) shown in FIG. 17 is a diagram showing a frequency domain in which the signal after interpolation is sampled again in S ′ units (S ′ <S). As shown in FIG. 17A, since the signal is a discrete signal, the signal is repeated in units of 1 / S 'in the frequency domain, but the band of the original signal is changed from -1 / 2S to 1
/ 2S, there is no information in the high frequency range.
In other words, the larger the relative value of the spatial distance between S and S ′, that is, the larger the enlargement ratio in the enlargement process, the more the difference is perceived. If resampling is desired, a signal in a band from -1 / 2S 'to 1 / 2S' is desired as shown in FIG. Of course, this is not the original signal sequence, but a signal that must be created during resampling.

For this reason, many techniques have been proposed for converting to higher image quality.

For example, US Pat. No. 5,280,546 discloses that a natural image and a line image are separated in order to remove interpolation blur.
Linear interpolation is applied to natural images, and linear interpolation is applied to line images.
An interpolation method for arranging the maximum value and the minimum value of the peripheral pixels by converting into a value is described. However, in this technology, interpolation blur occurs in a natural image, and since a high resolution information is created for a line image while reducing the resolution of the low image information, jaggies are unavoidable. There was a problem. In addition, since the line image is also quantized into two gradations, it is good if the original information has only two gradations. That was inevitable.

US Pat. No. 5,430,811 discloses a non-linear interpolation technique using a look-up table (LUT). However, in this technique, the algorithm itself only supports a magnification of 2 × 2 times, and in order to enlarge to another magnification, it is necessary to repeat this processing or use it together with another enlargement processing. There was a problem that the processing became complicated. Moreover, even if the power-of-two magnification is realized in the repetition processing, it is not possible to control each pixel of the high-resolution information to have a desired nonlinearity in the final magnification, due to the configuration of the repetition processing. It was not easy. Further, in this technique, since the pixel value of the observation pixel (original information) is not changed, the above-mentioned US Pat.
Jagging was inevitable, as was USP 5,280,546.

US Pat. No. 5,054,100 is also a proposal for nonlinear interpolation. However, this technique is effective for simple interpolation blur having edges in the vertical direction and the horizontal direction, but has a problem that the effect cannot be obtained if the image information becomes a little complicated. And like the above two proposals, the occurrence of jaggies was inevitable.

In addition, a binary filter that has been subjected to pseudo gradation processing such as a dither method or an error diffusion method for a long time is passed through a digital filter near a pixel of interest, converted to a multi-valued image, and then subjected to linear interpolation and re-encoding. There is a proposal to realize good resolution conversion by performing binarization (US Pat. No. 4,803,558, Japanese Patent Publication No. 61-61).
No. -56665). There is also a proposal to apply these techniques and perform a good resolution conversion process when the original information is a multi-valued image. However, as long as a quantization technique such as binarization is used in the method of creating an edge, the pixel value is not changed. It is difficult to form a good multi-valued image with spatial continuity.

SUMMARY OF THE INVENTION The present invention has been made to solve the above-mentioned problems, and an object of the present invention is to provide an image processing apparatus and method for converting input low-resolution image information into high-quality high-resolution image information. Aim.

[0024]

According to the present invention, there is provided an image processing apparatus for converting the resolution of input image information from a low resolution to a high resolution, the apparatus including a pixel of interest of the input image information. Detecting means for detecting a plurality of representative values from the pixel values of the peripheral pixels, interpolation value calculating means for obtaining an interpolated value of a plurality of interpolated pixels for interpolating the target pixel, based on an interpolation function, and the plurality of representative values; Calculating means for calculating output interpolated values of the plurality of interpolated pixels by a product-sum operation with interpolated values of the plurality of interpolated pixels.

According to another aspect of the present invention, there is provided an image processing method for converting the resolution of input image information from a low resolution to a high resolution, the method comprising the steps of: A detection step of detecting a plurality of representative values from the values; an interpolation value calculating step of obtaining an interpolation value of a plurality of interpolation pixels for interpolating the pixel of interest based on an interpolation function; Calculating an output interpolation value of the plurality of interpolation pixels by a product-sum operation with the interpolation value of the pixel.

[0026]

Embodiments of the present invention will be described below in detail with reference to the drawings.

[First Embodiment] FIG. 1 is a schematic block diagram showing the configuration of an image processing apparatus according to the first embodiment. The image processing apparatus of the present invention is efficiently provided mainly inside an image output apparatus such as a printer connected to a computer or a video printer for inputting a video signal, but an image processing apparatus other than the image output apparatus, for example, Needless to say, it may be incorporated as application software in the host computer or printer driver software for outputting to a printer.

As an example of the resolution conversion of the first embodiment, a case where input image information is converted into information having n times the number of pixels and m times the number of pixels will be described.

In the figure, reference numeral 100 denotes an input terminal to which image information in which one pixel is composed of low-resolution c bits (c> 1) is input. This low-resolution information is stored and held for several lines by the line buffer 101. And 1
In the windowing unit 02, a window including a plurality of peripheral pixels including the pixel of interest is set based on the image information for the several lines. This window is generally a rectangle centered on the pixel of interest, but it goes without saying that it may be other than a rectangle. Here, a description will be given assuming that the window size is a rectangle of 5 × 5 pixels as shown by the thick line in FIG. The target pixel is “M” in the figure.

Reference numeral 103 denotes a maximum value / minimum value detection unit which detects the maximum value (MAX) and the minimum value (MIN) in a 5 × 5 window (25 pixels of pixels A to Y) shown in FIG. To detect. An interpolation function calculation unit 104 interpolates one pixel of the target pixel into n × m pixels from the above-described window. A feature of the present embodiment is that the interpolation processing is interpolation using an interpolation function. The actual processing is based on the cubic convolution interpolation method (Cubic convolution interpolation) that approximates the interpolation function described in the conventional example, rather than the operation of the interpolation function.
It is more practical to implement the method by n) from the viewpoint of processing speed and the like.

Here, the interpolation value at each interpolation point (i, j) is set to p (i, j). Reference numeral 105 denotes a product-sum coefficient calculation unit, and MAX and M detected by the maximum value / minimum value detection unit 103.
Based on the value of IN, each product-sum coefficient of MAX, MIN, and p (i, j) is calculated. Now, let the product-sum coefficient of MAX be α, M
Let the product-sum coefficient of IN be β and the product-sum coefficient of P (i, j) be γ. In addition, the product-sum coefficient calculation unit 105 calculates each product-sum coefficient by evaluating the contrast value (referred to as CONT) calculated by the following equation.

CONT = MAX-MIN (Equation 5) As the value of CONT increases, the values of the product sum coefficients α and β are increased, and the value of the product sum coefficient γ is reduced. The method of calculating these product-sum coefficients will be further described later.

Next, reference numeral 106 denotes a product-sum operation unit which outputs the output value h at the interpolation point (i, j) based on the value of each product-sum coefficient.
(I, j) is calculated by the following calculation.

H (i, j) = αMAX + βMIN + γP (i, j) Expression 6 107 denotes an output terminal, and the calculated h (i, j) is output for n × m pixels per pixel of low-resolution information. Is done.

Next, the third-order convolution interpolation method will be described. Since this interpolation method has already been described with reference to FIG. 13, the details will be omitted. However, when the target pixel is M in the window shown in FIG. Change the pixel member of the operation. That is, interpolation is executed for each divided region by using 16 pixels surrounding the divided region.

FIG. 3 is a diagram showing a state in which the target pixel "M" of the window shown in FIG. 2 is divided into four parts. Now, each area is defined as M00, M10, M0 in the order of upper left, upper right, lower left, lower right.
1, M11. Specifically, as shown in FIG.
00 is A, B, C, D, F, G, H, I, K, L, M,
16 pixels of N, P, Q, R, S, M10 is B, C, D,
E, G, H, I, J, L, M, N, O, Q, R, S, T
16 pixels, and M01 is F, G, H, I, K, L, M,
16 pixels of N, P, Q, R, S, U, V, W, X, M1
1 is G, H, I, J, L, M, N, O, Q, R, S,
Interpolation is performed by 16 pixels of T, V, W, X, and Y. Naturally, as shown in FIG. 2, if there are 5 × 5 25 peripheral pixels, the third-order convolution interpolation method can be sufficiently realized.

Next, how to create high-resolution information from low-resolution information will be described. 5 and 6 are diagrams simply showing the idea of the present invention. For ease of explanation, 1
Expressed in dimensions.

In FIG. 5, the horizontal axis represents a one-dimensional coordinate space, the pixel of interest is j, and the left and right peripheral pixels are (j-2), (j-1), (j + 1), and (j + 2). And In the target pixel j, the number of pixels for which information is created by enlargement is referred to as a target pixel block. That is, the target pixel 1
The pixel values for the target pixel block are created from the pixels. The direction of the vertical axis indicates the depth direction of the pixel value, and points indicated by ● are the pixel values (observed values) of the original information of the target pixel and the peripheral pixels. In addition, the observed values are represented by I (j) and I (j), respectively.
-1), I (j + 1), I (j-2), I (j + 2). Then, a line connecting the pixel values of the original information of the target pixel and the surrounding pixels by an interpolation function (cubic convolution interpolation method) is defined as D (k) (k is a coordinate value by a spatial interpolation point). For ease of explanation, lines are connected in an analog manner.

The broken line is the average of the maximum value (MAX) and the minimum value (MIN) of the peripheral pixels, which is indicated by Ave. That is, Av
e is as follows.

Ave = (MAX + MIN) / 2 Expression 7 In this embodiment, the density transition line D generated by the interpolation function
By modifying (k), creation of a new high-resolution edge is realized by a continuous trajectory. D as described above
The density transition line (k) has only the Nyquist limit frequency band at the time of low input resolution. As the enlargement ratio increases, the gradient of the edge becomes gentler and is perceived as interpolation blur. Therefore, assuming now that an edge is formed by peripheral pixels and the pixel of interest falls inside the edge,
As shown in FIG. 6, the deformation is made such that the gradient is steep around the edge.

The deformation of the density transition line is calculated by extrapolating the density transition line of D (k) and the straight line of Ave. That is, assuming that a transition line that changes the slope steeply is h (k), the following equation is obtained. Here, a is a coefficient.

H (k) = a × D (k) − (a−1) × Ave Expression 8 By substituting Expression 7 into Expression 8, the following expression is obtained.

H (k) = a × D (k) − (a−1) × (MAX + MIN) / 2 Expression 9 Rearranging this results in the following expression.

H (k) = a × D (k) + (− (a−1) / 2)) MAX + (− (a−1) / 2)) MIN) Equation 10 That is, generally, The equation holds. Where α + β + γ =
It is one.

H (k) = αMAX + βMIN + γD (k) Expression 11 FIG. 6 is a diagram showing concentration transition lines of h (k) in two types of γ = 2 and 3. In each case, the slope of D (k) is steep centering on the straight line of Ave. That is, in other words, the spatial coordinate position of the center of the edge is represented by D
As in the case of (k), only the angle of the density transition is changed without changing.

FIG. 7 shows that α = −1 / 2, β = −1 / 2, γ
FIG. 9 is a diagram illustrating a state of edge creation when = 2. Note that h
Since the transition line in (k) increases the contrast,
It is necessary to set a limit depending on the density value. In the example shown in FIG. 7, the values of MAX and MIN are set as limit values. Further, a portion indicated by a bold line becomes new high-resolution information of the target pixel block. As a result, as is apparent from the figure, it is possible to realize the creation of an edge having a steep gradient.

At this time, attention must be paid to maintaining continuity when moving to the next target pixel. Even if the continuity is maintained in the target pixel block, if the density value is divided between the target pixel blocks, a pseudo contour may be generated due to discontinuity of gradation (tone jump). Therefore, as a countermeasure, detection of MAX and MIN
M from a window that completely contains one edge
A method of detecting AX and MIN, a method of fixing the value of Ave within a plurality of target pixels, or a method of generating target pixel block information of a plurality of target pixels instead of a target pixel block of a plurality of target pixels Various methods are conceivable, such as a method of creating group information. In any case, the idea of increasing the gradient of the interpolation function according to the present invention is not changed.

Next, a method for adjusting the product-sum coefficient according to the above-mentioned CONT will be described. FIG. 8 is a diagram showing an example in which the value of the coefficient α in Expression 11 is increased depending on CONT (β is β = α). (A) shown in FIG.
FIG. 8B shows an example in which CONT and α linearly change.
FIG. 8C shows an example in which it changes nonlinearly, and FIG. 8C shows an example in which it changes stepwise. Either example is possible, but (c) shown in FIG. 8 is the easiest and most realistic.

In this embodiment, the coefficient α increases as the CONT increases. That is, the contribution rate of the interpolation function is increased in the flat portion, which is a low frequency range, and the smooth signal continuity unique to the interpolation function is reproduced. On the other hand, in the edge portion, which is a high-frequency region, in order to widen the band to a further high-frequency region by resampling, the contribution ratio of the interpolation function is reduced, and the edge is sharply processed, thereby creating a new signal. In brief, the idea of the present embodiment lies in a configuration in which the advantages of the interpolation function are fully utilized, and the missing points are compensated for by processing.

As described above, Equation 10 or Equation 1
If a pixel value is calculated using a function based on 1 (Equation 6 in two dimensions), it is possible to create a good edge without blurring even when processing a natural image. Further, the adjustment of the coefficient can be realized by evaluating not only the CONT but also what kind of image based on the number of gradations of peripheral pixels, the distribution state of pixel values, and the like.

Although the approximation of the interpolation function is performed by using the third-order convolution interpolation method, a higher-order approximation may be used, or a method of calculating a SINC function may be used.

[Second Embodiment] Next, a second embodiment according to the present invention will be described in detail. The second embodiment further includes a plurality of sum-of-products operation units and an output value selection unit that selects an output from each of the sum-of-products operation units, and compares and selects several different interpolated values. High-resolution edge information is also freely estimated and created.

FIG. 9 is a schematic block diagram showing the configuration of the image processing apparatus according to the second embodiment. In FIG. 9, the same parts as those in FIG.

The low-resolution image information input from the input terminal 100 is stored in the line buffer 101 for several lines. Based on the image information for the several lines, the windowing unit 102 performs a process in units of windows using a plurality of peripheral pixels including the target pixel. Maximum / minimum value detector 1
03 detects MAX and MIN from the image information in the window. Then, the interpolation function calculation unit 104 performs the third-order convolution interpolation as in the above-described embodiment. 901
Denotes a product-sum coefficient calculation unit, which calculates coefficients of two combinations in this embodiment. One combination of coefficients is 9
02 is transmitted to the product-sum operation unit A and the other coefficient combination is transmitted to the product-sum operation unit B 903. Based on the values output from both the product-sum operation units, an output value selection unit 904 selects which operation result is finally output.

Next, the product-sum operation unit in this embodiment will be described. Information at the interpolation point k interpolated by the interpolation function is defined as D (k). First, the product-sum operation unit A902
Then, similarly to Equation 10, the converted density transition line hA (k)
In order to calculate the following, the following calculation is performed.

HA (k) = a × D (k) + (− (a−1) / 2)) MAX + (− (a−1) / 2)) MIN Equation 12 This is similar to FIG. , Ave, the calculation content of which changes the inclination around the straight line. That is, pixel values around the high-resolution edge to be created are formed.

The product-sum operation unit B 903 calculates the following 2
Two operations are performed. Note that the concentration transition line is represented by hB1 (k),
hB2 (k). Here, d and e are coefficients, and 0 <d <
1, 0 <e <1.

HB1 (k) = (d × D (k) + (1-d) × MAX) Expression 13 hB2 (k) = (e × D (k) + (1-e) × MIN) Expression 14 Here, the condition that the sum of the coefficients is 1 is satisfied in any of Expressions 12, 13, and 14. Further, the two values hB1 (k) and hB2 (k) form a pixel value in a portion that is slightly spaced from the center of the edge than the pixel value around the edge to be created. That is, it has the same role as the limit in FIG.

Next, the output value selection section 904 compares the values after the operation and selects the output value h (k) as follows.

When hA (k) ≧ hB1 (k), h (k) = hB1 (k) When hB1 (k)> hA (k) ≧ hB2 (k), h (k) = hA1 (k) When hB2 (k)> hA (k), h (k) = hA2 (k) Expression 15 FIG. 10 is a diagram showing a state of the output value selection unit 904 selecting the output value h (k). In the figure, the solid line part shown by the thick line is the high-resolution information h (k) created by the pixel block of interest.
Is equivalent to The coefficient a in hA (k) is a = 2
And the coefficient d in hB1 (k) is d = １ ／, hB
The coefficient e in 2 (k) is set to e = １ ／.

As is clear from the bold line, it is possible to change the inclination steeply at the edge portion of the high resolution to be created, and at a portion spatially separated from the center of the edge. Moreover, the change in the inclination can be realized with continuity. That is, any high-resolution edge information can be freely estimated and created by comparing and selecting several different values after interpolation.

Also in the present embodiment, as described above, it is necessary to maintain continuity with the previous target pixel block when the processing shifts to the next target pixel.

Although the creation of one-dimensional high-resolution edges has been described above, the same applies to two-dimensional expansion. A window for detecting MAX and MIN values;
The windows required for interpolation may be the same size or different.

Further, in the present embodiment, the output value is determined by comparing the magnitudes of the three density transition lines, but a comparison of a plurality of transition lines may be performed.

In the embodiment described above, the interpolation function calculation unit and the product-sum calculation unit are described separately for ease of explanation. However, the SINC function is also used in the interpolation function calculation unit as described above. It is needless to say that the interpolation function operation unit may be included in the product-sum operation unit if a method approximated by the product-sum operation of surrounding pixels is used.

As described above, according to the embodiment,
When converting the input low-resolution information into high-resolution information and during the enlarging process, it is possible to realize a conversion process with good image quality without blurring due to interpolation, which has been a particular problem in natural images. In spite of using the calculation by the interpolation function, since the information is created in the high frequency range equal to or higher than the frequency of the input information, it is extremely easy to create a natural and clear image with continuity even in a natural image. it can.

Further, since the conversion condition is dynamically changed based on the state of the partial pixel value of the low-resolution image, a good conversion having the interpolation function can be performed in the low-frequency region, and in the high-frequency region. Further higher image quality can be realized than the conversion using only the interpolation function.

Further, by making the edge portion sharp, occurrence of jaggies is reduced.

By making the algorithm of the present invention hardware or software, high quality output can be expected even for an image distributed on the Internet or an image with a small amount of information such as an image input from a digital video camera. Various products such as printers, video printers, and application software can be provided.

The present invention can be applied to a system including a plurality of devices (for example, a host computer, an interface device, a reader, a printer, etc.), and can be applied to a single device (for example, a copier, a facsimile). Device).

Further, an object of the present invention is to provide a storage medium storing a program code of software for realizing the functions of the above-described embodiments to a system or an apparatus, and to provide a computer (CPU or MP) of the system or apparatus.
It goes without saying that U) can also be achieved by reading and executing the program code stored in the storage medium.

In this case, the program code itself read from the storage medium implements the functions of the above-described embodiment, and the storage medium storing the program code constitutes the present invention.

As a storage medium for supplying the program code, for example, a floppy disk, hard disk, optical disk, magneto-optical disk, CD-ROM, CD
-R, a magnetic tape, a nonvolatile memory card, a ROM, or the like can be used.

When the computer executes the readout program code, not only the functions of the above-described embodiment are realized, but also the OS (Operating System) running on the computer based on the instruction of the program code. ) May perform some or all of the actual processing, and the processing may realize the functions of the above-described embodiments.

Further, after the program code read from the storage medium is written into a memory provided in a function expansion board inserted into the computer or a function expansion unit connected to the computer, based on the instruction of the program code, It goes without saying that the CPU included in the function expansion board or the function expansion unit performs part or all of the actual processing, and the functions of the above-described embodiments are realized by the processing.

[0076]

As described above, according to the present invention,
It is possible to convert the input low-resolution image information into high-quality high-resolution image information and output a natural and clear image with continuity even in a natural image.

[0077]

[Brief description of the drawings]

FIG. 1 is a schematic block diagram illustrating a configuration of an image processing apparatus according to a first embodiment.

FIG. 2 is a diagram illustrating an example of a window in the embodiment.

FIG. 3 is a diagram showing a state in which a target pixel “M” of the window shown in FIG. 2 is divided into four parts.

FIG. 4 is a diagram illustrating an example of interpolating a target pixel in a window illustrated in FIG. 2;

FIG. 5 is a diagram illustrating an interpolation process in the embodiment.

FIG. 6 is a diagram showing density transition lines of h (k) in two types of γ = 2, 3;

FIG. 7 is a diagram showing how an edge is created when α = −1 / 2, β = −1 / 2, and γ = 2.

FIG. 8 is a diagram illustrating an example in which the value of a coefficient α in Expression 11 is increased depending on CONT.

FIG. 9 is a schematic block diagram illustrating a configuration of an image processing apparatus according to a second embodiment.

FIG. 10 is a diagram showing how an output value selection unit 904 selects an output value h (k).

FIG. 11 is a diagram for explaining a nearest neighbor interpolation method in a conventional example.

FIG. 12 is a diagram for explaining a bilinear interpolation method in a conventional example.

FIG. 13 is a diagram for explaining a third-order convolution interpolation method in a conventional example.

FIG. 14 is a diagram illustrating an example of input information.

FIG. 15 is a diagram showing an example of processing by the nearest neighbor interpolation method.

FIG. 16 is a diagram showing an example of processing by bilinear interpolation;

FIG. 17 is a diagram for explaining the third-order convolution interpolation method in the frequency domain.

FIG. 18 is a diagram for explaining creation of high-frequency components in comparison with FIG. 17;

Claims (9)

[Claims] 1. An image processing apparatus for converting the resolution of input image information from a low resolution to a high resolution, comprising: detecting means for detecting a plurality of representative values from pixel values of peripheral pixels including a target pixel of the input image information. An interpolation value calculating means for obtaining an interpolation value of a plurality of interpolation pixels for interpolating the target pixel based on an interpolation function; and a product-sum operation of the plurality of representative values and the interpolation values of the plurality of interpolation pixels. An arithmetic unit for calculating output interpolation values of the plurality of interpolation pixels. 2. The method according to claim 1, wherein the interpolated value calculating means obtains interpolated values of the plurality of interpolated pixels by performing a multiply-and-accumulate operation of a product sum coefficient approximating a higher-order interpolation function and pixel values of peripheral pixels. The image processing device according to claim 1. 3. The image processing apparatus according to claim 2, wherein said interpolation value calculation means is a cubic convolution interpolation method. 4. The arithmetic means calculates a relative ratio between a product-sum coefficient of the plurality of representative values and a product-sum coefficient of an interpolated value of the plurality of interpolated pixels by a pixel value of a surrounding pixel including the target pixel. 2. The image processing apparatus according to claim 1, wherein the output interpolation values of the plurality of interpolation pixels are calculated by dynamically changing the interpolation values according to the state of (1). 5. The image processing apparatus according to claim 4, wherein the relative ratio changes according to contrast of surrounding pixels including the target pixel. 6. The image processing apparatus according to claim 5, wherein the relative ratio increases as the contrast increases. 7. A plurality of interpolation value calculating means for performing a product-sum operation of a product-sum coefficient of the plurality of representative values and a product-sum coefficient of an interpolation value of the plurality of interpolation pixels; 2. The image processing apparatus according to claim 1, further comprising a selection unit that selects an output interpolation value based on a calculation result by the unit. 8. The image processing apparatus according to claim 1, wherein the plurality of representative values are a maximum value and a minimum value in surrounding pixels including a target pixel. 9. An image processing method for converting input image information from low resolution to high resolution, comprising: detecting a plurality of representative values from pixel values of peripheral pixels including a target pixel of the input image information. An interpolation value calculating step of obtaining an interpolation value of a plurality of interpolation pixels for interpolating the target pixel based on an interpolation function; and a product-sum operation of the plurality of representative values and the interpolation values of the plurality of interpolation pixels. A calculating step of calculating output interpolation values of the plurality of interpolation pixels.