JP2002319020A - Image processing device and method - Google Patents

Abstract

PROBLEM TO BE SOLVED: To achieve the process of enlarging an image at high quality with small processing load and at high speed, without impairing the quality of the image, such as making its edges blurred or jagged. SOLUTION: This image processing device, which carries out an enlargement process for an original image through pixel interpolation, is provided with an edge detection means 40 for extracting edge components from the original image, an edge information conversion means 41 for producing edge strength information not including noise components, from the edges extracted, and an interpolation computing means 42 for making variable the shape of an interpolation function used in the enlargement process according to the edge strength information produced.

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 having a function of performing image enlargement processing. Here, the image enlargement process is a generic term for a process of adding new pixels to the pixels constituting the original image by interpolation, and a typical example is a case where the image size is enlarged (A4 size). → A
For example, there is a process required when a low-resolution image is output at a high resolution (eg, 400 dpi → 600 dpi).

[0002]

2. Description of the Related Art An image enlargement process is one of the basic processes for an image processing apparatus used in a system for editing, filing, displaying, printing, etc., an image. In particular, in recent years, with the spread of images on so-called Internet homepages and images corresponding to display resolutions taken by digital video, the number of printouts of these low-resolution images using a high-resolution printer has increased. Along with this, the importance of enlargement processing capable of realizing higher image quality is increasing.

In general, as an existing method used when performing image enlargement processing, for example, if a processing object is a color image expressed in multiple gradations (hereinafter, simply referred to as a “multi-valued image”), an original method is used. The nearest neighbor method using the pixel values of an image as they are, the linear interpolation method in which the change in value between pixels is assumed to be linear and linearly interpolating to obtain pixel values, and sinc based on the sampling theorem It is conceivable to use a cubic convolution method or the like for obtaining a pixel value to be interpolated using an interpolation function approximating the function (sin (x) / x). However, in each of these methods, jagged or mosaic-like degradation occurs in the enlarged image (nearest neighbor method), or the entire image is slightly blurred around the edge portion (linear interpolation method). ), It is difficult to achieve high image quality of the enlargement processing, such as light jaggies occurring at the edge portion or noise components being emphasized due to the characteristic of high frequency emphasis (cubic convolution method). There is.

From these facts, for example, Japanese Patent Application Laid-Open No.
4636, JP-A-7-129759, JP-A-6-54172, "High-quality image reconstructing by adaptive two-dimensional sampling function"
Vol. 8, No. 5, p. 620-626) and the like, a method of enlargement processing for solving the above-described difficulties is proposed.

[0005]

However, in these conventional enlargement processing methods, for the following reasons,
It cannot always be said that high-quality enlargement processing can be performed at high speed.

For example, Japanese Patent Laid-Open Publication No. Hei 7-114636 discloses that, as shown in FIG. 11, a pixel group PCU having an unknown pixel value is interpolated with respect to an original image, and for each pixel included in the pixel group PCU, If the number of neighboring pixels that are known is equal to or greater than the threshold value, an unknown pixel value is determined using the average value of the neighboring known pixels. It is disclosed that the processing speed is increased and the image quality is improved by obtaining the enlarged image finally by lowering and repeating the same processing. However, in such a method, since enlargement is performed by interpolation, the position of a known pixel remains as it is, and a position shift occurs in an image. Further, since the average value is used for determining the unknown pixel value from the known pixel group, there is a problem that the image is blurred. Therefore, even if high-speed processing is possible, it cannot be said that high-quality enlargement processing is realized.

For example, Japanese Patent Application Laid-Open No. 7-129759 discloses a JPEG (Joint Photographic Expert Group).
p) An inverse discrete cosine transform (hereinafter, referred to as “DCT”) for image data that has been subjected to a discrete cosine transform (hereinafter, abbreviated as “DCT”) by compression or the like by substituting “0” into a high-frequency portion that has not been used before In the technique of enlarging by "IDCT"), the function at the time of inverse transformation is converted into another function having a visually small error, instead of the IDCT, by the characteristic of the DCT coefficient or the user's instruction. It is disclosed that a small enlargement process is performed. However, in this method, since data subjected to DCT conversion is used for each block, enlargement at an arbitrary magnification cannot be performed. If the expanded DCT coefficient size is not a power of "2", there is a problem that the operation becomes complicated. Therefore, even if high image quality can be achieved, it cannot be said that high-speed and flexible enlargement processing is realized.

Japanese Patent Application Laid-Open No. 6-54172 discloses, for example, an enlargement by DCT / IDCT by a combination of DCT / IDCT and an iterative method of Gerchberg-Papoulis as shown in FIG. Discloses that it is possible to obtain a high-quality enlarged image with less blur and jaggy by estimating and restoring an unknown high-frequency component to which an appropriate value such as “0” is substituted.
However, in such a technique, D, which is a process with a large amount of computation,
Since it is necessary to perform CT / IDCT repeatedly on an image, an enormous amount of calculation time is required. Also, in the figure, there is a problem that magnification at an arbitrary magnification cannot be performed from the condition that "all of nN, mM, and nmN must be integer values". Therefore, even in such a method, high-quality image processing is possible, but high-speed and flexible enlargement processing cannot be said to be realized.

Further, "high-quality image enlargement reconstruction by an adaptive two-dimensional sampling function" is performed by using an approximation function truncated finitely because the sinc function is an infinite series in the cubic convolution method. Error and sampling theorem are intended for continuous and differentiable signals, whereas images have many discontinuous points. , A sinc function and an edge direction are detected globally and the interpolation function is deformed in that direction to obtain an enlarged image with less blur and jaggy. However, the function used in this method has a problem that the convolution matrix size is 4n with respect to the magnification n although it is local, and the amount of calculation is large due to detection of a global edge direction and the like.
That is, even with such a method, high image quality can be achieved, but high-speed enlargement processing cannot be said to be realized.

Accordingly, the present invention has been made in view of the above-described problems of the prior art, and is not limited to the case of performing enlargement processing on a multi-valued image expressed in multiple gradations, such as blurring of an edge portion and jaggies. It is an object of the present invention to provide an image processing apparatus and an image processing method capable of performing enlargement processing with high image quality without deteriorating image quality, and performing high-speed enlargement processing with a small processing load.

[0011]

SUMMARY OF THE INVENTION The present invention is an image processing apparatus devised to achieve the above object. That is, the present invention is an image processing apparatus for performing an enlargement process on an original image by interpolation of a new pixel, wherein the edge detection unit extracts an edge component in the original image, and a noise component is extracted from an extraction result by the edge detection unit. Edge information converting means for generating intensity information of an edge not including the edge information, and performing the enlarging processing by changing the shape of an interpolation function used in enlarging processing according to the edge intensity information generated by the edge information converting means. And an interpolation calculating means.

Further, the present invention provides an image processing method devised to achieve the above object, wherein an edge component in the original image is extracted when performing enlargement processing on the original image by interpolation of a new pixel. Generating intensity information of an edge that does not include a noise component from the extraction result, and performing the enlarging process by changing the shape of an interpolation function used in the enlarging process according to the generated edge intensity information. I do.

According to the image processing apparatus having the above configuration or the image processing method having the above procedure, the shape of the interpolation function is changed in accordance with the edge intensity information. Will be reflected. Therefore, even if the pixel to be interpolated is obtained from the neighboring pixels and the enlargement process is performed, it is possible to minimize the occurrence of image quality deterioration such as blurring of an edge portion or jaggy in the processed image. Moreover, since the interpolation pixel can be specified from the neighboring pixels,
The processing load does not become enormous.

[0014]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS An image processing apparatus and an image processing method according to the present invention will be described below with reference to the drawings.

[Explanation of Apparatus Configuration in this Embodiment] First, a schematic configuration of an image processing apparatus according to the present invention will be described. FIG. 1 is a block diagram illustrating a configuration example of a main part of an image processing apparatus according to the present invention, and FIG. 2 is a block diagram illustrating a schematic configuration example of the entire image processing apparatus according to the present invention.

As shown in FIG. 2, the image processing apparatus described here includes an image data input unit 10, an image data storage unit 11, an enlargement processing unit 12, an image data output unit 13,
It is provided with.

The image data input unit 10 receives image data of a multivalued image to be enlarged and has a function of communicating with an external source that supplies the image data. . The image data input to the image data input unit 10 includes JPEG format, BMP (bitmap) format, PNG that can be processed by the image processing apparatus.
(Portable Network Graphics), and more specifically, those described in an image format such as an application program that operates on a personal computer, and those captured by a digital camera.

The image data storage unit 11 has a function of temporarily storing image data input to the image data input unit 10 until the image data is output to the enlargement processing unit 12. A function of temporarily storing image data after resolution conversion until the image data is output to the image data output unit 13;
In the case where the image data input to the CPU is compressed, for example, it has a function of decompressing the compressed image data.

The enlargement processing section 12 includes an image data input section 10
After the image data is input to the image data storage unit 11, the image data received via the image data storage unit 11 is subjected to an enlargement process, and the image data is converted to a designated resolution. The details of the enlargement processing unit 12 will be described later.

The image data output section 13 receives the image data converted to the resolution after the enlargement processing from the image data storage section 11 and outputs the image data as enlarged image data. Communication function with an external device (for example, a printer).

Next, the main part of the image processing apparatus configured as described above, that is, the enlargement processing unit 12 for performing the image enlargement processing will be described in more detail. Enlargement processing unit 12
Is configured to include an edge detection processing unit 40, an edge information conversion processing unit 41, and a pixel interpolation processing unit 42, as shown in FIG.

The edge detection processing section 40 extracts an edge component from image data to be processed in the enlargement processing. Specifically, for example, using a local edge detection filter such as a Laplacian operator, a portion where the values of the constituent pixels of the image data rapidly change is extracted as an edge component.

The edge information conversion processing unit 41 estimates a multi-valued variable called a line process that is consistent with the image data to be processed from the edge components extracted by the edge detection processing unit 40, This is to be edge intensity information that does not include noise components related to image data.

The pixel interpolation processing unit 42 performs processing based on the multi-valued line process estimated by the edge information conversion processing unit 41.
The shape of the interpolation function used in the enlargement process is changed, and the pixel value to be interpolated is determined using the interpolation function after the change, whereby the enlargement process is performed on the image data to be processed.

[Explanation of Processing Operation in the Present Embodiment] Next, an example of a processing operation in the image processing apparatus configured as described above, that is, a processing procedure of the image processing method according to the present invention will be described.

When performing image enlargement processing, first, image data to be processed is input to the image data input unit 10, temporarily stored in the image data storage unit 11, and then transmitted to the enlargement processing unit 12. Sent. When the image data to be processed is received, in the enlargement processing unit 12, the edge detection processing unit 40 extracts an edge component from the image data. Up to this point, a known technique may be used, and therefore, details thereof are omitted here.

[Processing in Edge Information Conversion Processing Unit] After the extraction of the edge component, the edge information conversion processing unit 41
Generates edge intensity information from the edge component.
Here, the generation processing of the edge intensity information performed by the edge information conversion processing unit 41 will be described in detail.

In general, since a local filter such as a Laplacian operator is used as an extraction result of an edge component obtained by the edge detection processing unit 40, unlike a reliable edge information for image data, an edge component is originally used. In many cases, noise (noise) that should not be included is included, or edge components are cut off in the middle. Therefore, in the edge information conversion processing unit 41, when the edge detection processing unit 40 extracts the edge component, it estimates a line process that is likely edge information consistent with the input image data using energy function minimization, The noise component is removed from the extraction result by the edge detection processing unit 40.

FIG. 3 is an explanatory diagram showing a line process for a certain pixel fi, j. The line process is a variable indicating the presence of an edge component between pixels adjacent to each other. That is, for example, the horizontal line process hi, j in the figure is a variable indicating the existence of an edge between the pixel fi, j and the pixel fi + 1, j, and the vertical line process vi, j is the pixel fi, j and the pixel fi, j. This is a variable indicating the existence of an edge between fi, j + 1. Further, the value of each line process is a random variable taking a variable value from “0” to “1.0”.

The edge information conversion processing unit 41 minimizes the energy function represented by the following equation (1) with respect to the line process, which is the above-mentioned random variable, by using, for example, a statistical relaxation method. Estimate a likely line process for the image data to be processed.

[0031]

(Equation 1)

In the equation (1), li, j is the estimated horizontal and vertical line processes, and ei, j is the local Laplacian operator or the like for the input image data detected by the edge detection processing unit 40. The edge component information f'i, j obtained by applying the filter is a line process li, j
Are the values of the adjacent pixels sandwiching.

The first term in the equation (1) is a term that requires that the estimated line process li, j and the edge component information ei, j match. The second term is a term that requires matching between the difference between adjacent pixel values and the line process li, j. The difference between the pixel values is small or large, and the value of the line process li, j matches the density difference. , The energy of this term becomes smaller. The third term is a term representing the interaction of the penalty and the line process on the existence of an edge (discontinuity) in the image itself.
Note that α, β, and γ are weighting coefficients of the first, second, and third terms, respectively, and are arbitrary values. The third term is expressed by the following equation (2).

[0034]

(Equation 2)

V (h) and V (v) in equation (2)
Is calculated by the line process of each neighboring system. FIG. 4 is an explanatory diagram showing a line process of a neighboring system of each of the horizontal line process hi, j and the vertical line process vi, j. In equation (2), the first and second terms are terms that require that a line process in the same direction does not exist immediately next to the equation, and the third term is a term that gives a penalty to the line process itself. is there.

Further, in the equation (2), the fourth term is the energy of the nonlinear interaction which cannot be given in a quadratic form, and requires that the line process must have continuity. It is. FIG. 5 is an explanatory diagram showing a specific example of the relationship between the continuity of the line process and the energy. As shown in the figure, the weighting factors Cp, Cc
C1 and C1 are supplied with individual energies according to the relationship between the respective line processes (specific manner of continuity). In other words, the line process is continuous (the
If it has a continuation or a turn, a small energy is given, but the line process is interrupted in the middle (ending in the figure) or is immediately adjacent (para in the figure).
llel), T-shaped or cross-shaped, gives high energy.

That is, the edge information conversion processing unit 41 defines the energy based on factors that affect the edge strength, such as the density difference between the pixels sandwiching the edge and the continuity of the edge components, and the energy corresponding to the defined energy. The function is minimized by a statistical relaxation method, which estimates a likely line process.

Next, an outline of the processing procedure of the line process estimation will be described. FIG. 6 is a flowchart showing an outline of the flow of the line process estimation.

In estimating a probable line process, the edge information conversion processing unit 41 first sets the temperature parameter T in the statistical relaxation method to a sufficiently high temperature (step 60; hereinafter, step is abbreviated as "S"). . Then, as an initial line process, a horizontal line process h
i, j and the vertical line process vi, j from “0” to “1.
A value of "0" is arbitrarily set (S61), and for the current state line processes hi, j and vi, j at a certain position, the current state energy function Eline is obtained by using the above equations (1) and (2). (1) is calculated (S62).

Thereafter, the edge information conversion processing unit 41 changes the line process hi, j 'of the next state candidate which is randomly changed using a random number or the like with respect to the line process hi, j and vi, j of the current state. vi, j ', and for the line processes hi, j' and vi, j 'of these next state candidates,
The energy function Eline is calculated using the equations (1) and (2).
(L ') is calculated (S63).

The energy functions Eline (l) and Eline
After calculating (l ′), the edge information conversion processing unit 41 calculates
Using these and the temperature parameter T, a next state line process is determined (S64). Specifically, dU represented by the following equation (3) is calculated from the energy function Eline (l) of the current state, the energy function Eline (l ') of the next state candidate, and the temperature parameter T.

[0042]

(Equation 3)

At this time, if dU <0, the edge information conversion processing unit 41 updates the line process at a certain position with the next state candidate line processes hi, j 'and vi, j'. When dU ≧ 0, the following (4)
Calculate the state transition probability q represented by the equation.

[0044]

(Equation 4)

In the equation (4), Z is a constant for normalization. If dU ≧ 0, select ξ such that 0 ≦ ξ ≦ 1. Then, using the state transition probability q calculated by the equation (4), if ξ ≦ q, the line process at a certain position is changed to the next state candidate line process hi, j ′.
And vi, j ', and if ξ> q, the state is not updated and the line processes hi, j and vi, j in the current state remain.

When the current iteration number of the line process estimation reaches the preset iteration number (S65), the edge information conversion processing section 41 ends the line process estimation processing. However, if not reached, the temperature parameter T is lowered by a predetermined amount (S66),
The processing in each of the above-described steps (S62 to S65) is performed again. At this time, as the annealing schedule for lowering the temperature parameter T, for example, a schedule represented by the following equation (5) may be used.

[0047]

(Equation 5)

In the equation (5), C is an initial constant of annealing, α is an acceleration parameter of annealing, and i is the number of repetitions of the line process estimation at the present time. In equation (5), the acceleration parameter α is used.
The invention is not limited to this. For example, there is essentially no change even if α = 1 and no acceleration parameter is used.

As described above, the edge information conversion processing section 41
Then, the energy function represented by the equation (1) is minimized by using the statistical relaxation method of the above-described procedure, and a probable multi-valued line process is estimated, whereby the edge component extracted by the edge detection processing unit 40 is extracted. , A noise component is removed from the original image, and edge intensity information obtained by normalizing the edge intensity (probability specified from the density difference or the like) that originally exists in the original image is obtained.

Here, the case where the energy function is minimized using the statistical relaxation method has been described as an example. However, the minimization is not limited to this method. This may be performed using a deterministic method such as a method, or a minimum solution may be obtained by another optimal solution search method.

[Processing in Pixel Interpolation Processing Unit] After the generation of the edge intensity information, the pixel interpolation processing unit 42 subsequently processes the image data to be processed based on the multivalued line process as the intensity information. Is performed. Here, the enlargement processing performed by the pixel interpolation processing unit 42 will be described in detail.

In the enlargement processing, the pixel interpolation processing unit 42 first determines the pixels of the interpolation pixels used for the enlargement processing on the image data to be processed based on the multi-valued line process estimated by the edge information conversion processing unit 41. Determine the value. At this time, the pixel interpolation processing unit 42 determines the interpolated pixel value using the sigmoid interpolation function g (x). A sigmoid function is a solvable differentiable real function that is defined by all real inputs and takes positive derivatives everywhere. As its argument increases, certain fixed upper bounds (“0” and “ 1)), which rapidly asymptotic to a fixed finite lower bound as its argument decreases. The sigmoid interpolation function g (x) is shown in the following equation (6).

[0053]

(Equation 6)

In the equation (6), 1 is a line process value between pixels estimated by the edge information conversion processing unit 41. Therefore, when the sigmoid interpolation function g (x) represented by the equation (6) is illustrated, as shown in FIG.
When the value of the line process value l is large (for example, l =
0.8), the sigmoid interpolation function g (x) has a step-like shape, but when the value of the line process value l is large (for example, l = 0.2), the sigmoid interpolation function g (x)
(X) has a linear functional shape. As a result, even in the image data after the image enlargement processing, the edge information corresponding to the value of the line process 1 is held.

Here, as a specific example, a case where image data is enlarged five times will be described. FIG. 8 is an explanatory diagram showing an outline of a case where image data is enlarged five times, FIG. 9 is an explanatory diagram showing a specific example of determining an interpolated pixel value, and FIG. 10 is a supplementary explanatory diagram of a pixel area to be interpolated.

For example, a line process for a certain pixel fi, j and a line process in the vicinity thereof are shown in FIG.
Let us consider a case where the value is estimated as shown in FIG. At this time, between the pixel fi, j and the pixel fi, j + 1 and the pixel fi, j
Since the line process values between the pixel fi and the pixel fi + 1, j are respectively “0.8”, the sigmoid interpolation function g (x) expressed by the equation (6) can be calculated as shown in FIG. It becomes a shape close to a typical step function. Therefore, the values of the interpolated pixel groups of the area A1 and the area A2 shown in FIG. 8B are, for example, in the case of fi, j <fi, j + 1 (fi + 1, j), respectively.
As shown in (1), it is determined by the sigmoid interpolation function g (x) having a shape close to the step function.

In FIG. 8A, a pixel fi + 1, j
And pixels fi + 1, j + 1 and pixels fi, j + 1 and pixels fi + 1,
Since the line process values between j + 1 and “j + 1” are “0.2”, the sigmoid interpolation function g (x) has a shape relatively close to a linear function as shown in FIG. Therefore, the values of the interpolated pixel groups in the areas B1 and B2 shown in FIG. 8B are also determined by the sigmoid interpolation function g (x) having a shape close to a linear function, similarly to the above case. .

After determining the values of the interpolation pixel groups of the areas A1, A2, B1, and B2, the values of the interpolation pixel group of the area C surrounded by these pixel groups are determined. However, at this time, the interpolation pixel values of the regions A1, A2, B1, and B2 have already been determined. Therefore, as shown in FIG. 10, the region C is divided into regions (for example, a region C ′) including known pixel values at both ends, and the sigmoid interpolation function g ( x), the area A1,
Pixel values are determined in the same manner as the respective interpolated pixels in A2, B1, and B2. The line process value 1 in this case is determined from the line process value estimated by the edge information conversion processing unit 41. For example, in the case of FIG.
= 0.2 is determined linearly.

As described above, the pixel interpolation processing unit 42
The shape of the sigmoid interpolation function g (x) is varied stepwise or linearly in accordance with the line process value estimated by the edge information conversion processing unit 41, while using the sigmoid interpolation function g (x). Determine the interpolation pixel value,
By complementing the image data to be processed with the interpolated pixel values, enlargement processing is performed on the image data.

Here, for example, if the image data is enlarged five times, the area A (A) shown in FIG.
1 and A2), area B (B1 and B2), area C
It has been described that the interpolation processing is performed in the order of (including C ′). However, the interpolation processing is not limited to this order, and the interpolation processing is performed on the image data in the horizontal direction (the interpolation processing of the area A1 and the area B1). It is also conceivable to perform interpolation processing in the order of performing vertical interpolation processing (area A2, area C and area B2). Further, the interpolation processing of the area C is not limited to the above-described method. For example, the weighted average value is used without using the sigmoid interpolation function g (x) and using the values of the neighboring pixels for which the interpolation value has been determined. The interpolation value may be determined by a known method.

[Explanation of Operation and Effect in this Embodiment] As described above, according to the image processing apparatus and the image processing method described in this embodiment, sigmoid interpolation is performed according to the line process value which is the edge intensity information. Since the shape of the function g (x) is variable, its sigmoid interpolation function g (x)
The edge intensity information is reflected on the value of the pixel to be interpolated. That is, if the neighboring pixels of the pixel to be interpolated constitute an edge portion, an interpolated pixel value that does not obscure the edge portion is determined. An interpolated pixel value to be maintained is determined. Therefore, even if the pixel to be interpolated is obtained from the neighboring pixels and the enlargement process is performed, image quality deterioration such as blurring or jaggies at the edge portion, which greatly affects the image quality, can be suppressed as much as possible. It is possible to realize a complicated enlargement process. Moreover,
In the enlargement processing, the interpolation pixel can be specified from the neighboring pixels.
It is not necessary to repeat this step repeatedly, so that the processing load does not become enormous, and as a result, high-speed enlargement processing can be realized. In addition, since the interpolated pixel value is determined using the sigmoid interpolation function g (x), it is possible to cope with enlargement processing at an arbitrary magnification and to secure flexibility of enlargement processing.

Further, according to the image processing apparatus and the image processing method described in the present embodiment, the continuity of the edge component and the density difference between pixels sandwiching the edge are estimated in estimating the line process value as the edge intensity information. Since both are based, human visual characteristics are reflected in the estimation result of the line process value. Therefore, a portion that is difficult for human eyes to recognize as an edge is removed as a noise component, while a degradation in image quality such as blurring or jaggedness is reliably suppressed for an original edge portion. Image quality can be improved. Moreover,
The energy is defined based on both the continuity of the edge component and the density difference between the pixels sandwiching the edge, and the line process value is estimated by minimizing the energy. , While avoiding an enormous processing load. However, this can be said to be the same even when only one of the continuity of the edge component and the density difference between pixels sandwiching the edge is based.

[0063]

As described above, according to the image processing apparatus and the image processing method of the present invention, even if the enlargement process is performed on the multi-valued image expressed by the multi-gradation, blurring of the edge portion or blurring of the edge portion may be caused. Enlargement processing can be performed with high image quality without image quality deterioration such as jaggies, and the enlargement processing can be performed at high speed with a small processing load.

[Brief description of the drawings]

FIG. 1 is a block diagram illustrating a configuration example of a main part of an image processing apparatus according to the present invention.

FIG. 2 is a block diagram illustrating a schematic configuration example of an entire image processing apparatus according to the present invention.

FIG. 3 is an explanatory diagram illustrating a line process for a certain pixel.

FIG. 4 is an explanatory diagram illustrating a line process of a neighboring system of a horizontal line process and a vertical line process.

FIG. 5 is an explanatory diagram showing a specific example of a relationship between continuity of a line process and energy.

FIG. 6 is a flowchart showing an outline of a flow of line process estimation.

FIG. 7 is an explanatory diagram illustrating a shape example of a sigmoid function.

8A and 8B are explanatory diagrams showing an outline of a case where image data is magnified five times; FIG. 8A is a diagram showing a specific example of a line process value between pixels; FIG. It is a figure showing a specific example.

FIG. 9 is an explanatory diagram showing a specific example of determining an interpolation pixel value in the case of FIG. 8;

FIG. 10 is a supplementary explanatory diagram of a specific example of a pixel area to be interpolated in the case of FIG. 8;

FIG. 11 is an explanatory diagram showing an outline of an example of conventional pixel interpolation.

FIG. 12 is an explanatory diagram showing an outline of another example of the conventional image enlargement process.

[Explanation of symbols]

10: image data input unit, 11: image data storage unit, 1
2 ... Enlargement processing unit, 13 ... Image data output unit, 40 ... Edge detection processing unit, 41 ... Edge information conversion processing unit, 42 ... Pixel interpolation processing unit

Continued on the front page F term (reference) 5B057 AA20 CA08 CA12 CA16 CB08 CB12 CB16 CD06 CE03 DA17 DB02 DB09 DC16 5C076 AA21 AA32 BB04 BB15 5C077 LL02 MP01 MP08 NN02 NP01 PP03 PP20 PP43 PP65 TT02

Claims (7)

[Claims] 1. An image processing apparatus for performing an enlargement process on an original image by interpolating a new pixel, comprising: an edge detection unit for extracting an edge component in the original image; and a noise component from an extraction result by the edge detection unit. Edge information conversion means for generating intensity information of edges not included; interpolation for performing the enlargement processing by changing the shape of an interpolation function used in enlargement processing according to the edge intensity information generated by the edge information conversion means An image processing apparatus comprising: a calculation unit. 2. The method according to claim 1, wherein the edge information conversion unit generates the edge intensity information based on at least one of a continuity of an edge component extracted by the edge detection unit and a density difference between pixels sandwiching the edge. The image processing apparatus according to claim 1, wherein: 3. The edge information conversion means defines energy based on at least one of a factor of continuity of an edge component and a factor of a density difference between pixels sandwiching the edge, and minimizes the energy. The image processing apparatus according to claim 1, wherein edge intensity information is generated. 4. The image processing apparatus according to claim 3, wherein the edge information conversion unit minimizes the energy by using a statistical relaxation method. 5. The image processing apparatus according to claim 3, wherein the edge information conversion unit minimizes the energy using a steepest descent method. 6. The apparatus according to claim 1, wherein said interpolation operation means uses a sigmoid function as said interpolation function.
The image processing apparatus according to 2, 3, 4, or 5. 7. An image processing method for performing an enlargement process on an original image by interpolation of a new pixel, wherein an edge component in the original image is extracted, and edge intensity information containing no noise component is generated from the extraction result. An image processing method characterized in that the enlargement process is performed by changing the shape of an interpolation function used in the enlargement process according to the generated edge intensity information.