JP2001008027A - Image magnification device - Google Patents

Abstract

PROBLEM TO BE SOLVED: To obtain an image magnification device, which magnifies an original image in matching with an output device, that can magnify a sharp image without a fogged edge and flickering. SOLUTION: An input image adjustment means 11 adjusts an original image obtained by an image input means 10 so that number of desired horizontal and vertical pixels are halved, and a vertical edge generating means 100, a horizontal edge generating means 101 and an oblique edge generating means 102 generate edge images in 3 directions from the adjusted original image. A level-up means 103 applies inverse wavelet transform to the 3 images and the image adjusted by the input image adjustment means 11 by regarding them as 4 sub-band components of the wavelet transform image to form a magnified image with number of pixels as a multiple of 4.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an image enlarging apparatus capable of realizing a clear and high-quality image enlarging.

[0002]

2. Description of the Related Art In the fields of image databases, high-definition color printing, and the like, various high-quality image processing functions are required, and one of them is image enlargement. In addition to enlarging an image, in addition to a function of the image processing system, in order to match between media having different resolutions, for example, a low-resolution image input by an electronic still camera or the like is converted into a laser printer, an inkjet printer, or the like. This is an important technique when outputting to a high-resolution printer.

As a conventional image enlarging method, a method of simply interpolating between pixels has been adopted. As a typical interpolation method, a linear interpolation method of determining an interpolation value using a distance ratio of sample pixels close to an image to be interpolated as shown in FIG. And the nearest neighbor method.

[0004]

(Equation 1)

In equation (1), Da represents pixel data at point A in FIG. 17, Db represents pixel data at point B, Dc represents pixel data at point C, and Dd represents pixel data at point D in FIG. E is the interpolation point to be found.

However, in the above method, since the linear interpolation method suppresses the frequency characteristic of the pass band, the linear interpolation method is smoothed under the effect of applying a low-pass filter (low-pass filter; LPF), and the image is sharp. There is a drawback that blurred images lacking in the expression of details are likely to occur. Further, the nearest neighbor method has a drawback that distortion is likely to occur due to a large amount of high-frequency leakage, and the distortion appears as a mosaic or jaggy at an edge portion.

[0007] To solve the above problem,
A method has been proposed in which an image signal in a real space is converted into an image signal in a frequency space and then enlarged using an orthogonal transform such as FFT (fast Fourier transform) or DCT (discrete cosine transform). The intent of these methods is to restore high-frequency components lost at the time of sampling, and to estimate and restore detailed information and edge information of the image, thereby achieving higher image quality of an enlarged image.

FIGS. 18 and 19 show these conventional examples.
These are disclosed in JP-A-2-76472 or JP-A-5-16.
7920 is an example. FIG. 18 is a block diagram showing the configuration, and FIG. 19 is a diagram schematically showing processing steps. First, an original image (n × n pixels) in a real space as shown in FIG. 19A is subjected to orthogonal transformation to be converted into a component in a frequency space. FIG. 19B is an image in a frequency space that can be generated by arranging the positions of the coefficient components as pixel positions of the image. This processing is performed by the original image orthogonal transform unit 1800 in FIG. At this time, the image of the frequency space is originally obtained by arranging the frequency components according to each component position,
It can be represented by an n × n matrix. The matrix after this frequency conversion becomes low frequency components as it goes to the upper left, and becomes high frequency components as it goes to the right and down. Next, the “0” component embedding means 1801 multiplies the area converted to the image in the frequency space by s (sn in FIG. 19C).
× sn area) is prepared, and the area obtained by the above-described orthogonal transformation is provided in the low frequency component area in the sn × sn area.
The n × n frequency domain shown in (b) is copied, and “0” is interpolated in the remaining high frequency domain. Finally, an inverse orthogonal transform unit 1802 performs inverse orthogonal transform on the sn × sn frequency domain to obtain an image signal in the real space multiplied by s as shown in FIG. 19D. The estimated enlarged image is output.

[0009] In addition to the method of interpolating "0" into the high frequency component, a process of repeating the normal transform and the inverse transform of the image signal by using orthogonal transform (Gerchberg-Papoulis repetition), for example, as disclosed in JP-A-6-54172. ), A method of restoring high-frequency components has also been proposed. Also, as disclosed in Japanese Patent Application Laid-Open No. 8-294001, a method has been proposed in which an orthogonal transform component of an original image is embedded in a low-frequency region, and frequency information obtained based on a prediction rule prepared in advance is embedded in a high-frequency region.

[0010]

However, the method of embedding and enlarging "0" in the high-frequency region obtains an enlarged image better than the interpolation method between pixels such as the linear interpolation method or the nearest neighbor method. However, since the high-frequency components deleted during sampling have not been restored,
There has been a problem that a sufficiently clear image cannot be obtained.

A method for restoring high-frequency components in the process of repeating normal and inverse transforms of an image using orthogonal transforms is as follows.
To repeat the normal conversion and the reverse conversion, the number of computations increases,
There is a problem in processing speed. The operation amount of the orthogonal transformation and the inverse orthogonal transformation is not so problematic when the enlargement ratio s is not large, but when the enlargement ratio s is large, the operation amount of the inverse transformation becomes almost s × n with respect to the operation amount of the normal transformation. It will increase in proportion. In particular, in actual processing on a two-dimensional block, the amount of computation increases almost in proportion to the cube of s × n. Furthermore, when a color image signal is enlarged, enlargement processing is required for a plurality of color components, and further processing time is required. Further, in the case where the image to be enlarged has a low resolution, there is an example in which the high-frequency component is not sufficiently restored.

Japanese Patent Application Laid-Open No. 8-294001 is a method taking these problems into consideration, and devises processing time and restoration of high-frequency components. However, in order to embed frequency information obtained based on a prediction rule prepared in advance in the high frequency region of the original image, a rule between high frequency components and other regions is created based on a large number of images in advance. It is necessary to create an appropriate rule base, and there is a concern that if it is not possible to create an appropriate rule, a sufficient effect cannot be exerted.

Further, the size of an image is generally arbitrary, and the processing time becomes longer as the size required for the orthogonal transform becomes larger. Therefore, in general, orthogonal transform is not applied to the image size once at a time due to the problem of processing speed or the like, and the image is processed with a block size of about 4 to 16 pixels. At that time, there is also a problem that discontinuity (block distortion) between blocks of the completed enlarged image occurs at a boundary portion.

[0014]

In order to solve the above-mentioned problems, an image enlarging apparatus according to the present invention utilizes multi-resolution analysis in wavelet transform.

An image enlarging apparatus according to the present invention is an image enlarging apparatus for obtaining an enlarged image of Ln pixels × Lm pixels by enlarging an input original image of n pixels × m pixels. Input image adjusting means for adjusting to 2 pixels × Lm. / 2 pixels, and image enlarging means for generating an enlarged image by applying a method based on wavelet transform to the image adjusted by the input image adjusting means, It is characterized by the following.

The image enlargement apparatus specifically includes an image adjusted in accordance with the number of pixels of the enlarged image to be obtained, and a vertical edge image, a horizontal edge image, and an oblique edge image of the image adjusted by a wavelet. It is regarded as four sub-band images constituting the converted image, and inverse wavelet transform is performed on the sub-band image to obtain an enlarged image of a desired pixel size (first embodiment described later). See).

The above-mentioned image enlargement device is, specifically, obtained from a 1/4 size reduced image in a low frequency region of an image obtained by subjecting an input original image to a wavelet transform.
The relationship between one edge image and the remaining three sub-band images in the converted image is determined, and the related information is used to determine the vertical edge image and the horizontal direction of the image adjusted according to the number of pixels of the enlarged image to be determined. Each of the edge image and the oblique direction edge image is corrected.

Then, the adjusted image and the corrected three edge images are regarded as four sub-band images constituting the wavelet transform image, and this sub-band image is subjected to inverse wavelet transform, thereby obtaining a desired result. This is for obtaining an enlarged image of a pixel size (see a second embodiment described later).

The image enlargement device specifically includes one average edge image obtained from a 1/4 size reduced image in a low frequency range of a wavelet transform image of an input original image, The association between the remaining three subband images in the image is determined, and one average edge image obtained from the image adjusted according to the number of pixels of the enlarged image to be obtained is corrected using the relevant information.

Then, the adjusted image and the three images obtained by the correction are combined into a wavelet transform image.
By performing inverse wavelet transform on the assumption that there are two subband images, an enlarged image having a desired pixel size is obtained (see a third embodiment described later).

The image enlarging apparatus according to the present invention is an image enlarging apparatus for enlarging an input original image to obtain an enlarged image. By applying a method based on the wavelet transform to the target image, a target image enlarging unit that generates an enlarged image having four times the number of pixels, and an enlarged image obtained by the target image enlarging unit is set as an enlarged target image. Means for returning the processing to the target image enlarging means; enlarged image presenting means for visually presenting the enlarged image obtained by the target image enlarging means; and enlarging processing for the enlarged image presented by the enlarged image presenting means. Alternatively, there is provided an image adjusting means for performing a reduction process, and an enlarged image output means for outputting an image obtained by the image adjusting means.

Specifically, when the size of the image to be enlarged is unclear and cannot be determined, the above-described enlargement method using the wavelet transform is applied. The enlarged image obtained by the inverse wavelet transform is used as the next target image. Then, the inverse wavelet transform is performed in order to repeat the enlargement processing to an image having four times the number of pixels. Then, the operation is stopped when the visually desired enlargement ratio is attained, thereby enlarging the input original image (see a fourth embodiment described later).

Also, the image enlargement device of the present invention is inputted.
Enlargement processing of a color digital image of n pixels × m pixels and Ln
In a color image enlarging apparatus that obtains an enlarged image of pixels × Lm pixels, a basic component selecting unit that selects a basic color component from the color image constituent components, and other color components from the basic color component selected by the basic component selecting unit. Conversion ratio deriving means for deriving a conversion ratio when deriving, a basic component image adjusting means for adjusting a basic color component of an input original image to Ln / 2 pixels × Lm. / 2 pixels, and a previous basic component image The adjusting means applies a method based on a wavelet transform to the basic color components of the adjusted image to generate an enlarged image, and an enlarged image of the basic color components obtained by the basic image enlarging means as a desired image. Basic enlarged image adjusting means for adjusting to Ln pixels × Lm pixels, and applying the conversion ratio of the conversion ratio deriving means to the enlarged image of the basic color component derived by the basic enlarged image adjusting means Deficient component enlarging means for estimating an enlarged image in another color component, an enlarged image of a basic color component derived by the basic enlarged image output means, and a remaining color component estimated by the deficient component enlarging means. And an enlarged image output means for generating an enlarged image of the color original image by synthesizing the image data.

More specifically, in the enlargement method using the wavelet transform, when the input image is a color image, a basic color component is determined from the color components constituting the input color image, An enlarged image for the basic color component is generated. The remaining color components are obtained by estimating the enlarged image of the basic color components by performing a linear conversion based on the conversion ratio, thereby improving the efficiency of the process of generating the enlarged image of the color image (fifth embodiment described later). Form).

[0025]

Embodiments of the present invention will be described below with reference to the drawings. FIG. 1 is a configuration diagram of an image enlargement device according to a first embodiment of the present invention, and FIG. 7 is a configuration diagram of an image enlargement unit configuring an image enlargement device according to a second embodiment of the present invention. FIG. 10 is a block diagram of each of the large image units constituting the image enlargement apparatus according to the third embodiment of the present invention, and FIG. 13 is a block diagram of the image enlargement apparatus according to the fourth embodiment of the present invention. FIG. 15 shows a configuration diagram of an image enlargement device according to a fifth embodiment of the present invention.

FIG. 4 is a schematic diagram of the processing steps of the image enlarging apparatus according to the first embodiment of the present invention, and FIG. 8 is a diagram illustrating the processing of the image enlarging apparatus according to the second embodiment of the present invention. FIG. 11 is a schematic diagram of the process, FIG. 11 is a schematic diagram of the processing steps of the image enlargement device according to the third embodiment of the present invention, and FIG. 14 is the process of the image enlargement device according to the fourth embodiment of the present invention. FIG. 16 is a schematic diagram of a process, and FIG. 16 is a schematic diagram of a process of an image enlargement apparatus according to a fifth embodiment of the present invention. In the drawings of the configuration, the same parts are denoted by the same reference numerals.

(First Embodiment) First, the first embodiment of the present invention will be described.
An image enlarging apparatus according to the embodiment will be described. FIG. 1 shows the configuration of an image enlargement device according to the first embodiment. In FIG. 1, reference numeral 10 denotes an input device typified by a scanner, a digital still camera, or the like, and image input means which is a means for reading digital images obtained from these input devices, and 11 denotes an image input means. Input image adjustment means for adjusting the number of horizontal and vertical pixels of an original image having n pixels × n pixels so as to be a half of a desired enlarged image size of Ln pixels × Ln pixels. Image enlarging means 13 for enlarging the used image is an Ln pixel × Ln pixel having a desired enlarged image size of 13 times the enlarged image having four times the number of pixels of the original image adjusted by 11 obtained by the image enlarging means 12. The enlarged image adjusting means 14 is an enlarged image output means for displaying the enlarged image data estimated at 13 and outputting it to another device or the like.

The image enlarging means 12 includes a vertical edge generating means 100 for extracting an edge component image in the vertical direction from the original image adjusted to Ln / 2 pixels × Ln / 2 pixels by the input image adjusting means 11, and an input image adjusting means. A vertical edge generating unit 101 for extracting an edge component image in the horizontal direction from the original image adjusted to Ln / 2 pixels × Ln / 2 pixels by the unit 11; and an Ln / 2 pixel × Ln / 2 pixel by the input image adjusting unit 11. A vertical edge generating means 102 for extracting an edge component image in an oblique direction from the adjusted original image, three edge component images obtained by 100 to 102, and Ln / 2 pixels × Ln / 2 pixels by the input image adjusting means 11 The adjusted original image is regarded as four of the sub-band components constituting the transformed image when the enlarged image of Ln pixels × Ln pixels is subjected to the wavelet transform, and an enlarged image of Ln pixels × Ln pixels by inverse wavelet transform Generate a And a level-up means 103 for performing the operation.

In the image enlargement apparatus according to the first embodiment configured as described above, multi-resolution analysis using wavelet transform is used for image enlargement. For wavelet transform, see “Wavelet Beginner's Guide”
(Susumu Sakakibara, Tokyo Denki University Press) and many other documents. The wavelet transform has been developed as a method of frequency analysis, and has been applied in many fields such as signal processing and data compression of images and the like.

When a wavelet transform is performed on an original image to obtain a wavelet transform image, it is composed of several subband components. FIG. 3 is a diagram showing an example of the arrangement of the subband components of the wavelet transform image, and the original image is represented by LL3, HL3, LH3, HH3, HL2, LH2, HH2, HL1, LH
1 shows an example of dividing into ten subband components of HL1.

FIG. 2 is a diagram in which the three-stage wavelet transform as shown in FIG. 3 is represented by a filter sequence. As is apparent from FIG. 2, the wavelet transform is independent in the vertical direction (y direction) and the horizontal direction (x direction). It is composed of a combination of processing Low for performing downsampling to 1/2 after low-pass filtering and processing High for downsampling to half after high-pass filtering. First, set the original image to High horizontally.
Low is performed. The result of performing High processing in the vertical direction with respect to the output when performing High processing in the horizontal direction is HH
One component, and the result of low filter processing in the vertical direction is HL
One component. Next, the result of low-level filtering in the vertical direction with respect to the output of the low-level filtering in the horizontal direction is LL1, and the result of high-level filtering in the vertical direction is the LH1 component. The above is the result obtained by the first wavelet transform.

Further, Low filter processing and High filter processing are performed on the LL1 component in the horizontal direction, and
The result of vertical filtering of the output when the gh processing is performed is HH2, and the result of low filtering in the vertical direction is HL2. In addition, the result of performing low filter processing in the vertical direction with respect to the output when performing low filter processing in the horizontal direction is LL2, and the result of performing high filter processing in the vertical direction is the LH2 component. The above is the result obtained by the second wavelet transform.

Similarly, for the LL2 component,
By performing the filtering process and the High filtering process separately, and also performing the Low filtering process and the High filtering process separately in the vertical direction, the subband components of HH3, HL3, LH3, and LL3 are obtained. The conversion has been performed.

As described above, the original image is decomposed into the four frequency components LL1, HL1, LH1, and HH1 by the first-stage wavelet transform.
Since the subsampling of / 2 is performed, the size of the image representing each component is １ ／ of the original image. LL1 is a low-frequency component extracted from the original image, which is a blurred image of the original image, but most information of the original image is included here. Therefore, in the next second stage wavelet transform, this LL1 is regarded as a target image.

On the other hand, the HL1 component represents an image in which high frequency components in the horizontal direction are strongly extracted in the original image by the filtering process in FIG. 2, and LH1 represents an image in which high frequency components in the vertical direction are strongly extracted in the original image. It will be. HH1 is an image in which high-frequency components are extracted in both the horizontal and vertical directions. In other words, HH1 can be considered as an image in which high-frequency components are extracted in oblique directions.

Considering this from the pixel value, the HL1 component strongly reflects an area where the pixel value varies greatly in the horizontal direction (edge information pointing in the vertical direction) in the original image.
LH1 strongly reflects an area in the original image where the pixel value changes greatly in the vertical direction (edge information pointing in the horizontal direction), and HH1 shows an area in which the pixel value changes greatly in the horizontal and vertical directions (the oblique direction). This is considered to largely reflect the (edge information pointing toward).

The above-described feature of the wavelet transform is also applicable to the components of LL2, HL2, LH2, and HH2 obtained by the second-stage wavelet transform for LL1, and similarly, for LL2. The components of wavelet transform as an image also hold. As described above, the wavelet transform converts the LL image obtained by extracting the low-frequency component in the sub-band image of the previous stage into four １ ／ resolution images corresponding to the low-frequency component, frequency components in the vertical, horizontal, and oblique directions. Can be considered to be decomposed into Then, by synthesizing these sub-band component images by filter processing, it is possible to restore the previous image. Considering FIG. 3, four subband images LL3 and HL
3, LL2 is restored by combining LH3 and HH3, LL2, HL
2, LL1 is restored by combining LH2 and HH2. Then, the original image can be restored by using LL1, HL1, LH1, and HH1. Thus, in the wavelet transform,
Since a plurality of sub-band component images having different resolutions can be simultaneously expressed, it is also called multi-resolution analysis, and it is noted that by compressing each sub-band component, efficient data compression can be performed. In the present invention, the original image is regarded as the low-frequency sub-band component LL1 in the previous stage, and
Image with strong extraction of the remaining horizontal high-frequency components
HL1, image L with strong extraction of high-frequency components in the vertical direction
It is applied to enlarge the original image to a desired image size by obtaining a double enlarged image by estimating H1, and an image HH1 in which high-frequency components are extracted in both the horizontal and vertical directions. .

With the above in mind, the operation of the first image enlarging device according to the present invention will be described. First,
First, in the image input means 10, n pixels to be enlarged × n
An original image having the size of a pixel is read. In this case, an input device such as a digital still camera or a scanner may be prepared and input data from the input device may be used, or a device for reading from a digital data file or the like from another device may be used. The original image obtained in 10 is subjected to image size adjustment in input image adjustment means 11. This is because, as described in the multi-resolution analysis in the wavelet transform, when the target image is once subjected to the wavelet transform, the subband components after the conversion are always １ ／ in size both in the number of horizontal pixels and the number of vertical pixels before the conversion. . Conversely, the pixel size obtained by the inverse wavelet transform has twice the number of horizontal pixels and the number of vertical pixels of the sub-band component at the time of performing the processing, that is, four times the number of pixels. From the above properties, it is desirable that the number of horizontal and vertical pixels of the enlarged pixel obtained by the inverse wavelet transform is a multiple of two. Therefore, in 11, first, Ln pixels × Ln pixels, which is a desired enlarged image size, are adjusted to be dLn pixels × dLn pixels, which is a multiple of 2, and the original image is adjusted to dLn pixels.
Adjust so that / 2 pixels x dLn / 2 pixels. At this time, various methods can be used as an adjustment method. Here, it is assumed that the adjustment is realized by interpolating between pixels using (Equation 1) or by thinning out pixels in an area where pixel variation is small. The original image is transformed into a frequency domain by an orthogonal transformation such as DCT transformation, and a frequency component corresponding to dLn / 2 pixels × dLn / 2 pixels is extracted, or a high-frequency component deficient with 0 is filled in, and it is converted to dLn /
Although it is possible to adjust by the inverse orthogonal transform corresponding to 2 pixels × dLn / 2 pixels, it is complicated here if considering the efficiency of processing and further expanding by the method by wavelet transform. Since the processing does not seem to be very efficient, simple pixel interpolation or thinning was adopted.

The image enlarging means 12 enlarges the original image adjusted to dLn / 2 pixels × dLn / 2 pixels by the input image adjusting means 11 twice in both the horizontal and vertical directions to obtain desired Ln pixels × Ln pixels. Is enlarged to an image size close to. At this time, the multi-resolution analysis in the wavelet transform described above is used. In this way, in the conventional method of performing orthogonal transformation and compensating for the shortfall in the frequency domain, there is a problem that jagged noise occurs at a joint of blocks by dividing the processing time into a plurality of blocks, The wavelet transform has an advantage that such a noise does not occur because a large image can be handled at once.

At 12, the sub-band component LL in FIG.
Assuming that the image is adjusted to dLn / 2 pixels x dLn / 2 pixels as 1, an image of dLn / 2 pixels x dLn / 2 pixels corresponding to the remaining three subband components of HL1, LH1, and HH1 is estimated. There is a need to.

FIG. 5 schematically shows this state. Here, a method is adopted in which three images HL1, LH1, and HH1 are regarded as edge images in three directions of LL1. As described above, the HL1 component is obtained by strongly extracting the high-frequency component in the horizontal direction from the enlarged image (referred to as LL0) having four times the number of pixels of the original image adjusted to dLn / 2 pixels × dLn / 2 pixels. LH1 represents an image in which high-frequency components in the vertical direction are strongly extracted in LL0. HH1 is an image in which high-frequency components are extracted in both the horizontal and vertical directions. That is, HL1 is considered to reflect a portion representing a high frequency component in the horizontal direction, that is, the edge information in the vertical direction in the LL0 image. On the other hand, LH1 is considered to reflect a portion representing a high frequency component in the vertical direction, that is, the edge information in the horizontal direction in the LL0 image. And HH1
Is considered to reflect a portion representing a high frequency component both vertically and horizontally in the LL0 image, that is, reflecting edge information in an oblique direction. Therefore, according to the present invention, the vertical edge generation unit 100 adjusts at 11
An edge component in the vertical direction of the original image adjusted to dLn / 2 pixels × dLn / 2 pixels is extracted and regarded as a missing HL1 component. Further, the horizontal edge generation means 101 extracts an edge component in the horizontal direction of the original image adjusted to dLn / 2 pixels × dLn / 2 pixels adjusted in 11, and removes the missing LH1.
Consider as a component. Similarly, in the oblique edge generating means 102,
An edge component in the oblique direction of the original image adjusted to dLn / 2 pixels × dLn / 2 pixels adjusted in 11 is extracted, and is regarded as a missing HH1 component. At this time, a filter for detecting three directions as shown in FIG. 6 is used as the edge detecting means. Note that this filter is not uniquely determined, and other filters are also possible. The level up unit 103 performs an inverse wavelet transform on the four subband components estimated as described above,
Obtain a clear enlarged image with the size of dLn pixels x dLn pixels. At this time, since the wavelet transform can be represented by filter series processing as shown in FIG. 2, the processing here may be a filter series processing for performing the reverse.

If the size of the enlarged image obtained in step 12 is not a desired image size of Ln pixels × Ln pixels, which is not a multiple of 2 (the value of Ln / 2 is not an integer), a slight shift occurs. Therefore, the enlarged image adjustment means 13 thins out the pixel interpolation so as to compensate for the slight deviation. Since the processing performed here is at most one pixel, the effect is small by processing at a portion where the change is small in the image (the gradation change is small and other than near the edge).

The enlarged image obtained through the enlarged image adjusting means 13 is displayed at 14 on a CRT or output so as to be input to another device.

As described above, according to the embodiment,
It is not only possible to reduce the edge blur caused by simply interpolating between pixels in the original image, or to fill in insufficient frequency components with 0 or the like in the frequency domain as in the conventional device, and it is also a problem in the orthogonal transformation method. A sharp enlarged image can be realized without generating noise such as jaggies. Also, it is possible to easily estimate without creating a rule or the like in advance.

(Second Embodiment) Next, a second embodiment of the present invention will be described.
An image enlarging apparatus according to the embodiment will be described. FIG. 7 shows a configuration of an image enlarging unit in the image enlarging apparatus according to the second embodiment. FIG. 8 schematically shows a processing process in the image enlarging apparatus according to the second embodiment.

In FIG. 7, reference numeral 700 denotes a wavelet transform for the original image adjusted to dLn / 2 pixels × dLn / 2 pixels by the input image adjusting means 11, and the obtained sub-band components are converted before the transform. The number of pixels is 1/2 in both the horizontal and vertical directions
, Input fine adjustment means for adjusting dLn / 2 pixels × dLn / 2 pixels to be a multiple of 2;
Wavelet transform is performed on the original image of ddLn pixels × ddLn pixels obtained in 00, and the four sub-band components LL2, HL2, LH2, and HH2 are level-down means for generating, and 702 is obtained in 701. Reference HL component generation means for detecting edge information in the vertical direction from the LL2 component in the low frequency region by edge filter processing and setting it as a reference HL component HLe among the subband components, and 703 denotes the subband obtained at 701 Among the components, reference LH component generation means for detecting edge information in the horizontal direction from the LL2 component in the low frequency range by edge filter processing to obtain a reference LH component LHe, 704 denotes a reference LH component of the subband component obtained in 701 Among them, a reference HH component generation unit that detects edge information in an oblique direction from the LL2 component in the low frequency range by edge filter processing and sets it as a reference HH component HHe. And 705 is 702
From the correlation between the reference HL component HLe obtained in step 701 and the HL2 component obtained in step 701, ddLn obtained by the vertical edge generation means 100
HL correction estimating means for obtaining a correction component dHL for correcting the HL component image HL1 with respect to the original image of pixels × ddLn pixels, and reference numeral 706 denotes a reference LH component LHe obtained at 703 and 7
LH correction estimating means for obtaining a correction component dLH for correcting the LH component image LH1 with respect to the original image of ddLn pixels × ddLn pixels obtained by the horizontal edge generating means 101 from the correlation of the LH2 components obtained in step 01. Reference numeral 707 denotes ddLn pixels × dd obtained by the oblique edge generation means 102 based on the correlation between the reference HH component HHe obtained in 704 and the HH2 component obtained in 701.
HH correction estimating means for obtaining a correction component dHH for correcting the HH component image HH1 with respect to the original image of Ln pixels;
08 is the correction component dHL obtained in 705 as ddLn pixels × ddLn
HL component estimating means for estimating the HL component of the original image of ddLn pixels × ddLn pixels from the HL component image HL1 obtained by the pixel-adjusted and HL component images obtained by the vertical edge generating means 100,
Reference numeral 709 denotes the correction component dLH obtained in step 706 as ddLn pixels × ddL.
LH component estimating means for estimating the LH component of the original image of ddLn pixels × ddLn pixels from the LH component image LH1 obtained by the horizontal edge generating means 101 and the LH component image obtained by the horizontal edge generating means 101; The component dHH adjusted to ddLn pixels × ddLn pixels and the oblique edge generating means 102
HH component estimating means for estimating the HH component of the original image of ddLn pixels × ddLn pixels from the HH component image HH1 obtained in the step (1).

The operation of the image enlarging apparatus according to the second embodiment configured as described above will be described with reference to the schematic diagram of the processing steps in FIG.

As in the image enlarging apparatus according to the first embodiment of the present invention, the original image obtained in step 10 is processed by the input image adjusting means 11 to obtain the desired enlarged pixel size Ln pixels × Ln pixels horizontally. The number of vertical pixels is adjusted to dLn pixels × dLn pixels, which are multiples of two, and the original image is adjusted to dLn / 2 pixels × dLn / 2 pixels, both vertically and horizontally having a size of １ ／. At 700, the level down means 701 further obtains a subband component one level lower from the original image adjusted to dLn / 2 pixels × dLn / 2 pixels in response to the result of 11, so that dLn / 2 pixels × dLn / Two
Fine adjustment is made to ddLn pixels × ddLn pixels by interpolating or thinning out one pixel so that the pixel is a multiple of two. And 70
In 1, a wavelet transform is performed on the original image of ddLn pixels × ddLn pixels obtained in 700, and four subband components LL2, HL2, LH having an image size of １ ／ of this are obtained.
2. Generate HH2.

FIG. 9 schematically shows an outline of the processing in the image enlarging means. In the case of the first embodiment of the present invention, the vertical edge image, the horizontal edge image, and the oblique edge image of the current target image LL1 are set to the current target image LL1.
Although the sub-band components HL1, LH1, and HH1 that were missing in the wavelet transform image of the image LL0 that was doubled were considered, this does not hold strictly with the filtering process of FIG. For example, when HL0 to HL1 are obtained by filter processing, high-frequency component data in the horizontal direction and low-frequency component data in the vertical direction are extracted as HL1 components from the filter processing of FIG. From this, HL1 components include:
A pixel portion having a large value variation in the horizontal direction (an edge portion extending in the vertical direction, etc.) and a pixel portion having a small value variation in the vertical direction are extracted. In the first embodiment of the present invention, it is considered that the influence of the pixel portion having a large value variation in the horizontal direction, that is, the edge portion extending in the vertical direction is large, and only the edge information in the vertical direction is HL.
Although considered as one component, the effect of a pixel portion with small value fluctuation in the vertical direction may not be ignored depending on the image to be handled. Further, the edge information in the vertical direction includes many pixel portions having a large value fluctuation in the horizontal direction, but does not exactly match. The same can be said for other LH1 and HH1. From the above, in the present invention, the current target image is once subjected to wavelet transform to lower the level by one to create the sub-band components LL2, HL2, LH2, and HH2, and the low-frequency component LL2 in the sub-band components From the correlation between the edge images HLe, LHe, HHe in the three directions and the actual three subband components HL2, LH2, HH2, the corresponding subband components are estimated from the three edge images of the original target image. The correction amounts dHL, dLH, and dHH were determined.

First, at 702, the LL2, HL2, LH2, HH2
LL2, which is in the low frequency range and represents the original image best
Paying attention to the components, edge information in the vertical direction is detected by a filter as shown in FIG. Refer to this HL component HL
e. Then, the HL estimating means 705 checks the correlation between the reference HL component HLe and HL2 obtained in 701. Although various methods are conceivable, here, a difference image dHL = HLe-HL2 between the reference HL component HLe and the actual HL component HL2 is determined as shown in FIG. 9C. Similarly, in reference numerals 703 and 704, the reference LH component LHe
And the horizontal edge information of the LL2 component is selected as the reference HH component HHe.
The H estimation means 706 obtains a difference image dLH = LHe-LH2 between the reference LH component LHe and the actual LH component LH2, and the HH estimation means 707 calculates the reference HH component HH
Difference image between e and actual HH component HH2 dHH = HHe-HH2
I asked. Then, in 708, 709, and 710, as in the first embodiment of the present invention, the vertical edge generating means 100
By adding the correction components dHL, dLH, and dHH corresponding to the HL1 obtained in the above, the LH1 obtained by the horizontal edge generation unit, and the HH1 obtained by the diagonal edge generation unit 102, the ddLn pixel × ddLn pixel is increased by 700. LL of the sub-band component of the adjusted original image
Estimate HL1, LH1, and HH1 components when 1 is considered. In 708 to 710, when using the correction components dHL, dLH, and dHH obtained in 705 to 707, the distance between each pixel is calculated according to (Equation 1) such that the pixel size of each correction image is ddLn pixels × ddLn pixels. It was adjusted by interpolation, but this is not unique. In addition to this, there is a conventional method of expanding the pixel size to twice in both the horizontal and vertical directions by filling the missing component with 0 in the frequency-converted area. Applicable. But,
Considering the efficiency of the processing and the fact that this sub-band component is used to perform enlargement processing by inverse wavelet transform, the effect on the image quality of the finally obtained enlarged image even if pixel interpolation is simply performed Since 小 さ い is small, it is assumed here that a method by linear interpolation such as (Equation 1) is used for 708 to 710.

The processing after step 103 is performed in the same manner as in the first embodiment of the present invention, and the enlarged image finally obtained by the enlarged image output means 14 is displayed on a CRT or input to another device. Is output as

HL and L in 708 to 710
As the H and HH component estimation, in addition to the addition of the difference component between the reference component and the actual component as described above, (1) the result of multiplying the conversion coefficient matrix in the correction amount obtained in 705 to 707 by 100 (2) Add the result obtained by converting the correction amount obtained in 705 to 707 by the conversion function to the result of 100 to 102. (3) Input the results of 705 to 707 and 100 to 102 (4) Input the results of 705 to 707 and 100 to 102 and output a large number of databases and rule bases prepared in advance by using the learned neural network model to output the estimated values of HL, LH, and HH. It is also possible to use a method such as estimating the HL, LH, and HH components from Eq.

By doing so, the missing sub-pixels in the wavelet transform image that could not be extracted by simply detecting the edges in the three directions of the simply adjusted original image taken in the first embodiment of the present invention. Band components, especially high-frequency components, can be estimated more accurately, and image blur can be reduced. Further, by using the wavelet transform, there is no need to perform block division as in the orthogonal transform, and block distortion, which has been a problem in the conventional orthogonal transform method, does not occur.

(Third Embodiment) Next, an image enlarging apparatus according to a third embodiment of the present invention will be described. FIG.
0 represents the configuration of the image enlarging means in the image enlarging device according to the third embodiment. FIG. 11 schematically shows processing steps of the image enlargement apparatus according to the third embodiment of the present invention.

In FIG. 10, reference numeral 1000 denotes an edge generating means for detecting an edge image from the original image obtained by the input fine adjusting means 700 by Laplacian filter processing;
Reference numeral 1001 denotes a reference that generates a reference component for correcting the estimated values of the HL1, LH1, and HH1 components corresponding to the original image obtained in 700 by Laplacian filtering from the LL2 component image obtained in the level down unit 701. It is a component generation means.

The operation of the image enlarging apparatus according to the third embodiment configured as described above will be described with reference to the schematic diagram of the processing steps in FIG. The processing in the image input means 10 and the input image adjusting means 11 is the same as that in the first embodiment of the present invention, and the processing in the input fine adjusting means 700 is as described in the second embodiment of the present invention. At 11,
From the original image adjusted to dLn / 2 pixels × dLn / 2 pixels, in order to obtain a subband component one level lower by the level-down means 701, dLn / 2 pixels × dLn / 2 pixels must be a multiple of 2. Then, fine adjustment is made to ddLn pixels × ddLn pixels by interpolating or thinning out one pixel.

In the reference component generation means 1001, the sub-band component image obtained in 701 is in the low frequency range.
From LL2, a reference component for obtaining each correction amount used for estimation in 708 to 710 is obtained.
Here, by using a Laplacian filter as shown in FIGS. 12A and 12B for LL2, an average edge image of LL2 is regarded as a reference component image. The Laplacian filter is often used as a method for detecting an edge that is hardly influenced by directionality, instead of detecting an edge in a specific direction as shown in FIG. In the present embodiment, this is used as a reference component image to determine the correction amount as performed in the second embodiment. By doing so, frequent edge detection processing, which has been a problem in the second embodiment, can be reduced, leading to more efficient processing. HL correction estimating means 70
5. LH correction estimating means 706, HH correction estimating means 707
Then, as in the second embodiment, a difference image dHL2 between the edge image HE obtained in 1001 and the HL2 obtained in 701,
Edge image HE obtained in 1001 and LH obtained in 701
The difference image dLH2 obtained from the second difference image dLH2, 1001 and the difference image dHH2 obtained from the HH2 obtained at 701 are obtained (Equation 1).
Each difference image is adjusted to an image having ddLn pixels × ddLn pixels by the linear approximation as described above. On the other hand, the edge generation means 100
At 0, an edge image is detected by a Laplacian filter from the original image of ddLn pixels × ddLn pixels adjusted at 11, and H
L component estimation means 708, LH component estimation means 709, HH
In the component estimating means 710, the sub-band components HL1, LH1,
HH1 is accurately estimated. By doing this, (2
× ddLn) pixels × (2 × ddLn) pixels can be clearly estimated, and the desired image can be obtained by adding interpolation or thinning of several pixels.
An enlarged image having Ln pixels × Ln pixels can be obtained with high accuracy.

As described in the second embodiment of the present invention, the correction in 708 to 710 may be performed in addition to the result obtained by multiplying the conversion coefficient matrix in the correction amount in 705 to 707 or the conversion function. It is also possible to use the transformed result.

As described above, in this embodiment, the second image enlarging device adjusts the vertical edge image of the adjusted original image,
The point using the horizontal edge image, the corrected image of the oblique direction edge image as a shortage subband component of the wavelet transform image with respect to the input original image, and the horizontal and vertical positions in the low frequency range of the wavelet transform image of the input original image. Investigate the relationship between one average edge image from a reduced image of １ ／ size in both vertical pixels and the remaining subband components, and apply it to the estimation of three subband components missing from the original image. In order to speed up the processing.

(Fourth Embodiment) An image enlarging apparatus according to a fourth embodiment of the present invention will be described. FIG. 13 shows a configuration of an image enlargement device according to the fourth embodiment, and FIG. 14 shows processing steps in the image enlargement device according to the fourth embodiment.

The gist of the present invention is that the number of pixels to be enlarged is not known in advance, and an enlarged image is generated based on the original image while doubling both in the horizontal and vertical directions according to a multiresolution analysis by wavelet transform. By presenting to the user, it is determined whether an appropriate enlarged image has been obtained.

In FIG. 13, reference numeral 1300 denotes an enlargement processing initializing means for setting an image to be enlarged.
The original image input at 0 is set. 1301 is 130
Target image enlarging means for enlarging an image of m pixels × m images set as an image to be enlarged at 0 to 2 m pixels × 2 m pixels using multi-resolution analysis by wavelet transform;
Reference numeral 2 denotes an enlarged image presenting unit for presenting the enlarged image obtained in 1301 to the user, and 1303 determines whether the enlarged image size presented to the user in 1302 is appropriate according to an instruction input from the user. A multiplex processing end determination unit 1304 for allowing the user to determine whether fine adjustment of the size of the enlarged image in 1303 is necessary,
An image fine adjustment unit as an image adjustment unit that performs fine adjustment of the enlargement processing or reduction processing of the enlarged image in 1303 when receiving an instruction that the processing is necessary.

The operation of the image enlarging apparatus according to the fourth embodiment configured as described above will be described with reference to the processing steps shown in FIG. The original image of n pixels × n pixels input at 10 is first set as an enlargement target image at 1300. Next, the target image enlarging means 1301 enlarges the original image of n pixels × n pixels to an image size twice as large in both the horizontal and vertical directions, ie, four times. This enlargement processing is the first processing of the present invention.
By applying the image enlarging means described in the image enlarging device of the third embodiment, the image can always be enlarged to a size having four times as many pixels as the target image. 1302 is 13
01 is a means for presenting the current enlarged image obtained in step 01 to the user through the CRT. If the resolution exceeds the CRT resolution, the function of moving the viewpoint with a cursor or the like is provided.
Providing the function of cutting out a specific portion helps the user to determine the appropriateness of the current enlarged image. At 1303, the instruction result from the user is received,
If it is appropriate, the process proceeds to the next image fine adjustment means 1304. If it is not appropriate, the enlarged image is set as the next target image and the process returns to 1301.

The image fine adjustment means 1304 confirms whether the user needs fine adjustment. Since this uses multi-resolution analysis by wavelet transform for image enlargement, it can always be generated only with an image size four times that before enlargement. Therefore, when viewed by the user, the previous image is still small, but this may not be drawn on the CRT at one time and may be considered to be too large. So 13
At 04, the user is again asked to make some adjustments to this image, and if it is to be slightly larger, pixel interpolation is performed. On the other hand, if the size is to be slightly reduced, the image size is adjusted again by thinning out the pixels. When performing the pixel interpolation, an area other than the edge where the fluctuation of the pixel value is small is selected. The same applies to thinning. However, it is also possible to take a method of converting the data into a frequency domain and adding a missing component or extracting a surplus component from a high frequency.
It is conceivable to take an appropriate method according to the U ability and the like. Reference numeral 14 performs output processing such as displaying the enlarged image obtained in 1304 on a CRT, outputting the image using a printer, or passing it as data to another device.

As described above, the fourth image enlarging apparatus according to the present invention presents the obtained fine enlarged image to the user to determine whether the size and resolution are appropriate,
Since it is sufficient to give an instruction to stop the series of enlargement work at the time when it is obtained, it is not necessary to determine the enlargement ratio of the enlarged image in advance, and it is possible to easily enlarge the image to a size desired by the user.

(Fifth Embodiment) Finally, the fifth embodiment of the present invention
An image enlarging apparatus according to the embodiment will be described. FIG. 15 shows the configuration of an image enlargement device according to the fifth embodiment, and FIG. 16 shows a flowchart of processing in the image enlargement device according to the fifth embodiment. When estimating an enlarged image of a color original image, there is a drawback that the enlargement processing is enormous compared to the enlargement processing of one-color multi-tone data. The present invention is an invention related to the improvement of efficiency.

In FIG. 15, 1500 is a basic component selecting means for selecting a basic color component from a color original image, 1501 is a conversion ratio deriving means for calculating a simple ratio between the basic color component and the remaining color components, 1502 Is n
Basic component image adjustment means for adjusting the number of horizontal and vertical pixels of the original image of the basic color component image having pixels × n pixels to a half of a desired enlarged image size of Ln pixels × Ln pixels. A basic image enlargement unit 1504 for performing image enlargement using analysis on the basic color component. Reference numeral 1504 denotes an enlarged image of the basic color component obtained by the basic image enlargement unit 1503 having a desired enlarged image size.
Basic enlarged image adjustment means for adjusting to Ln pixels × Ln pixels, 15
Reference numeral 05 denotes insufficient component enlargement means for estimating the enlarged image of the remaining color components from the enlarged image of the basic color component obtained in 1504, and 1506 re-enlarges the enlarged image of each color component obtained so far into one. This is an enlarged color image reconstructing means to be constituted.

The operation of the image enlarging apparatus according to the fifth embodiment configured as described above will be described with reference to the processing steps shown in FIG. A basic color component is selected at 1500 from the color original image input at 10. Normally, the structure of a color original image is composed of three color components of red, green, and blue. Considering that green data is largely reflected on luminance information, green data is considered to be appropriate as a basic color component. But,
Other colors are acceptable. It is also possible to convert the three signals of red, blue and green into color data of another color system and select a basic color component therefrom. For example, it is possible to convert to luminance, hue, and saturation and to select luminance as the basic color component.However, here, the red, blue, and green color system is used, and the basic color component is green. Data will be used.
The conversion ratio deriving means 1501 obtains a simple ratio ratio # r of the red component to the green data and a simple ratio ratio # b of the blue component to the green data. There are various methods for calculating the ratio. Here, as shown in (Equation 2), the average value of the ratio of red to green in the target region and the ratio of blue to green in the target region are calculated. The average value of the ratio will be used.

[0069]

(Equation 2)

In (Equation 2), r # ij is red data at the pixel position (i, j) in the original image, g # ij is green data at the pixel position (i, j) in the original image, b #ij is blue data at the pixel position (i, j) in the original image. However, instead of estimating all the remaining components expanded using one ratio coefficient in this way, a matrix R # r composed of a ratio of red component to green data at each pixel and a blue component for green data It is also possible to adopt a matrix R # b composed of a simple ratio of In this case, it is possible to reproduce the characteristics of the original color original image rather than using one ratio coefficient, and it is possible to enlarge the color image with higher accuracy. At this time, since the number of elements of the conversion matrix obtained from the color original image is smaller than the number of pixels for the enlarged image, it is necessary to interpolate the missing part.

Basic image enlargement processing is performed on the green data, which is the basic color component, at 1503 in the same manner as in the image enlargement apparatuses according to the first to third embodiments of the present invention.
The obtained enlarged image of the basic color component is adjusted by pixel interpolation or thinning so that the desired image size becomes Ln pixels × Ln pixels in 1504. Then, the shortage component enlarging means 1505 applies the simple ratio ratio # r of the red component to the green data obtained in 1501 and the simple ratio ratio_b of the blue component to the green data to the expanded green data 1504. Then, enlarged data for the remaining red and blue components is created. By combining these three enlarged components into one, an enlarged image of the color original image can be obtained. In 14, the enlarged image is displayed on a CRT or the like, passed to an output device such as a printer, or sent to another image processing device. The output processing is performed with the data handled in.

By performing such processing, it is not necessary to perform enlargement processing for each of a plurality of components constituting the color original image, so that the processing can be simplified and speeded up.

[0073]

As described above, the first image enlarging device according to the present invention comprises input image adjusting means for adjusting an input original image to Ln / 2 pixels × Lm. / 2 pixels, and the input image adjusting means. And an image enlarging means for generating an enlarged image by applying a method based on the wavelet transform to the image adjusted by the above, so that the edge of the enlarged image can be generated without causing block distortion generated by the conventional orthogonal transformation method. It becomes possible to easily reduce rattling and blurring.

The image enlarging device of the first embodiment is
The relationship between the three edge images obtained from the quarter-size reduced image in the low frequency range in the wavelet transform image for the input original image and the remaining three subband images in the transform image will be described. Edge detection by correcting each of the vertical edge image, horizontal edge image, and oblique edge image of the original image adjusted according to the number of pixels of the enlarged image obtained using the related information. Insufficient high-frequency components in a wavelet transform image that could not be extracted by performing the above operation can be more accurately estimated.

Further, the image enlarging device according to the second embodiment
A vertical edge image, a horizontal edge image, and a diagonal edge image of the adjusted original image are obtained and regarded as insufficient subband components of a desired enlarged image, and the original image is also corrected for those subband components. Are obtained by using a vertical edge image, a horizontal edge image, and a diagonal edge image of a quarter-size reduced image obtained by performing the wavelet transform on.

The image enlarging apparatus according to the third embodiment estimates an insufficient subband component of a desired enlarged image using an average edge image obtained by one Laplacian filter and further performs a wavelet of the original image. By performing correction using the converted image, the processing speed has been further increased.

The image enlarging apparatus according to the present invention is an image enlarging apparatus for enlarging an input original image to obtain an enlarged image. By applying a method based on the wavelet transform to the image, a target image enlarging unit that generates an enlarged image having four times the number of pixels, and an enlarged image obtained by the target image enlarging unit is set as an enlarged target image. Means for returning the processing to the image enlarging means; enlarged image presenting means for visually presenting the enlarged image obtained by the target image enlarging means; and enlarging processing or enlarging the enlarged image presented by the enlarged image presenting means. The image processing apparatus includes: an image adjustment unit that performs a reduction process; and an enlarged image output unit that outputs an image obtained by the image adjustment unit.

According to this image enlargement apparatus, the enlargement method using the above wavelet transform is applied when the enlargement ratio of the enlarged image is unknown, and the enlarged image obtained by the inverse wavelet transform is An enlarged image can be sequentially generated as an image and presented to the user. Then, when the user obtains an image size (image resolution) that the user prefers based on the presented image, the user may instruct stop of a series of enlargement work, and performs image enlargement by trial and error. There is no need for it, making it a simpler user interface.

Further, the image enlargement device of the present invention
Enlargement processing of a color digital image of n pixels × m pixels and Ln
In a color image enlarging apparatus that obtains an enlarged image of pixels × Lm pixels, a basic component selecting unit that selects a basic color component from the color image constituent components, and other color components from the basic color component selected by the basic component selecting unit. Conversion ratio deriving means for deriving a conversion ratio when deriving, a basic component image adjusting means for adjusting a basic color component of an input original image to Ln / 2 pixels × Lm. / 2 pixels, and a previous basic component image The adjusting means applies a method based on a wavelet transform to the basic color components of the adjusted image to generate an enlarged image, and an enlarged image of the basic color components obtained by the basic image enlarging means as a desired image. Basic enlarged image adjusting means for adjusting to Ln pixels × Lm pixels, and applying the conversion ratio of the conversion ratio deriving means to the enlarged image of the basic color component derived by the basic enlarged image adjusting means Deficient component enlarging means for estimating an enlarged image in another color component, an enlarged image of a basic color component derived by the basic enlarged image output means, and a remaining color component estimated by the deficient component enlarging means. And an enlarged image output means for generating an enlarged image of the color original image by synthesizing the image data.

According to this image enlarging apparatus, when the input image is a color image, the reference color component is determined from the color components constituting the input color image,
An enlarged image for the basic color component is generated. Then, for the remaining color components, by performing linear conversion on the enlarged image of the basic color component by the conversion ratio and estimating, it is possible to increase the efficiency of processing in generating an enlarged image of a color image.

[Brief description of the drawings]

FIG. 1 is a block diagram illustrating a configuration of an image enlargement device according to a first embodiment of the present invention.

FIG. 2 is a diagram illustrating a filtering process in a wavelet transform.

FIG. 3 is a diagram schematically illustrating subband components in a wavelet transform image.

FIG. 4 is a diagram schematically showing processing steps of the image enlargement apparatus according to the first embodiment of the present invention.

FIG. 5 is a diagram schematically illustrating estimation of an insufficient subband component in a wavelet transform image.

FIG. 6 is a diagram illustrating an example of a filter applied to edge detection in the vertical, horizontal, and oblique directions.

FIG. 7 is a block diagram illustrating a configuration of an image enlargement unit included in an image enlargement device according to a second embodiment of the present invention.

FIG. 8 is a diagram schematically illustrating processing steps of an image enlargement apparatus according to a second embodiment of the present invention.

FIG. 9 is a diagram schematically illustrating estimation and correction of an insufficient subband component in a wavelet transform image.

FIG. 10 is a block diagram illustrating a configuration of an image enlargement unit included in an image enlargement apparatus according to a third embodiment of the present invention.

FIG. 11 is a diagram schematically illustrating processing steps of an image enlargement apparatus according to a third embodiment of the present invention.

FIG. 12 is a diagram illustrating an example of a Laplacian filter applied to edge detection.

FIG. 13 is a block diagram illustrating a configuration of an image enlargement device according to a fourth embodiment of the present invention.

FIG. 14 is a diagram schematically illustrating processing steps of an image enlargement apparatus according to a fourth embodiment of the present invention.

FIG. 15 is a block diagram illustrating a configuration of an image enlargement device according to a fifth embodiment of the present invention.

FIG. 16 is a diagram schematically illustrating processing steps of an image enlargement apparatus according to a fifth embodiment of the present invention.

FIG. 17 illustrates a conventional linear interpolation method.

FIG. 18 is a block diagram illustrating a configuration of a conventional image enlargement device.

FIG. 19 is a view for explaining an example of conversion to a conventional frequency domain and enlargement;

[Explanation of symbols]

 Reference Signs List 10 image input means 11 input image adjustment means 12 image enlargement means 13 enlarged image adjustment means 14 enlarged image output means 100 vertical edge generation means 101 horizontal edge generation means 102 oblique edge generation means 103 level up means 700 input fine adjustment means 701 level down Means 702 Reference HL component generation means 703 Reference LH component generation means 704 Reference HH component generation means 705 HL correction estimation means 706 LH correction estimation means 707 HH correction estimation means 708 HL component estimation means 709 LH component estimation means 710 HH component estimation means 1000 Edge generation means 1001 Reference component generation means 1300 Enlargement processing initialization means 1301 Target image enlargement means 1302 Enlarged image presentation means 1303 Multiple processing end determination means 1304 Image fine adjustment means 1500 Basic component selection means Step 1501 Conversion ratio deriving means 1502 Basic image adjusting means 1503 Basic image enlarging means 1504 Basic magnifying image adjusting means 1505 Missing component enlarging means 1506 Enlarged color image reconstructing means 1800 Original image orthogonal transform means 1801 "0" component embedding means 1802 Inverse orthogonal Conversion means

 ────────────────────────────────────────────────── ─── Continuation of the front page (72) Inventor Akio Kojima 1006 Kazuma Kadoma, Kazuma, Osaka Prefecture Matsushita Electric Industrial Co., Ltd. F term (reference) 5B057 AA11 CA16 CB01 CB16 CD06 CD07 CD10 CE03 CE06 CE16 CG10 CH09 DB06 DC16 5C076 AA21 AA31 BA02 BA06 BB06 BB07 BB42 CA12 CB01

Claims (11)

[Claims] 1. An image enlarging device for enlarging an input n-pixel × m-pixel original image to obtain an enlarged image of Ln-pixels × Lm-pixels, wherein the input original image is Ln / 2 pixels × Lm. / An image enlargement unit comprising: an input image adjustment unit that adjusts to two pixels; and an image enlargement unit that generates an enlarged image by applying a method based on a wavelet transform to the image adjusted by the input image adjustment unit. apparatus. 2. The image enlarging means includes: a vertical edge generating means for generating a vertical edge image of the image adjusted by the input image adjusting means; and a horizontal edge of the image adjusted by the input image adjusting means. Vertical edge generating means for generating an image, oblique edge generating means for generating an edge image signal in an oblique direction of the image adjusted by the input image adjusting means, and a vertical edge image obtained by each of the edge generating means;
The horizontal edge image, the oblique edge image, and the original image adjusted by the input image adjusting means are regarded as subband components in the wavelet transform image, and the inverse bandlet transform is performed on the subband component to increase the number of pixels by four times. The image enlargement apparatus according to claim 1, further comprising: a level-up unit that generates an enlarged image having the following. 3. An image enlarging unit, wherein the input fine adjusting unit adjusts again the number of horizontal and vertical pixels of the image adjusted by the input image adjusting unit so that the number of pixels becomes a multiple of two. Level-down means for performing a wavelet transform on the adjusted image; reference HL component generation means for generating a vertical edge image from a sub-band component located in a low frequency band of the transformed image obtained by the level-down means; Reference LH component generation means for generating a horizontal edge image from a subband component located in a low frequency range of the converted image obtained by the level down means; low frequency range of the converted image obtained by the level down means A reference HH component generating means for generating an edge image in a diagonal direction from the sub-band component located at HL correction amount estimating means for indicating a relation with a sub-band component belonging to a vertically low frequency region of the image obtained in the step, and an image of the reference LH component generating means and an image of the image obtained by the level down means. LH correction amount estimating means for indicating a relationship with a subband component belonging to a low frequency region in the horizontal direction, and a subband located in a high frequency region of the image of the reference HH component generating means and the image obtained by the level down means HH correction amount estimating means representing the relationship with the component, vertical edge generating means for generating a vertical edge image of the adjusted original image, and H vertical edge image obtained by the vertical edge generating means as H
An HL component estimating means for correcting the result of the L correction amount estimating means into an HL sub-band component, a horizontal edge generating means for generating a horizontal edge image of the adjusted original image, and a horizontal edge generating means. Let the obtained horizontal edge image be L
HL component estimating means for correcting the result of the H correction amount estimating means into LH sub-band components, oblique edge generating means for generating an oblique edge image of the adjusted original image, and oblique edge generating means The obtained oblique edge image is represented by H
HH component estimating means for correcting the result of the H correction amount estimating means into HH subband components; HL obtained by the LH, HL, HH component estimating means
The three images of the LH and HH components and the adjusted image are regarded as subband components in the wavelet transform image, and inverse wavelet transform is performed on the subband component to generate an enlarged image having four times the number of pixels. 2. The image enlarging apparatus according to claim 1, further comprising a level-up means for performing the function. 4. An image enlarging unit, wherein the input fine adjusting unit adjusts again so that the number of horizontal and vertical pixels of the image adjusted by the input image adjusting unit becomes a multiple of two, and the input fine adjusting unit adjusts the image. Level down means for performing a wavelet transform on the obtained image; reference component generation means for obtaining a Laplacian edge image from a subband component located in a low frequency band of the image obtained by the level down means; HL correction amount estimating means for indicating a relationship between the obtained reference component image and a sub-band component belonging to a vertically low frequency region of the converted image obtained by the level down means; An LH correction amount estimating method representing the relationship between the reference component image obtained and the sub-band component belonging to the horizontal low-frequency region of the converted image obtained by the level down means. HH correction amount estimating means for indicating a relationship between a reference component image obtained by the basic image generating means and a sub-band component in a high frequency range of the image obtained by the level down means; An edge generating means for generating a Laplacian edge image; an HL component estimating means for correcting the edge image obtained by the edge generating means according to the result of the HL correction amount estimating means to obtain an HL sub-band component; An LH component estimating unit that corrects the obtained edge image based on the result of the LH correction amount estimating unit to obtain an LH subband component, and corrects the edge image obtained by the edge generating unit based on the result of the HH correction amount estimating unit. HH component estimating means for obtaining the HH subband component, and HL obtained by the LH, HL, HH component estimating means,
The three images of the LH and HH components and the adjusted image are regarded as subband components in the wavelet transform image, and inverse wavelet transform is performed on the subband component to generate an enlarged image having four times the number of pixels. The image enlarging apparatus according to claim 1, further comprising: a level-up means for performing the function. 5. When the values of Ln / 2 and Lm./2 are integers, the input original image is adjusted by the input adjustment means, and the adjusted image is enlarged by the image enlargement means. When an image is generated and at least one of Ln / 2 and Lm./2 is not an integer, the input original image is adjusted by the input adjustment means, and the adjusted image is enlarged by the image enlargement means. The image enlargement apparatus according to claim 1, wherein an enlarged image is generated by the image enlargement unit, and the enlarged image obtained by the image enlargement unit is adjusted to Ln pixels x Lm pixels by the enlarged image adjustment unit. 6. An image enlarging apparatus for enlarging an input original image to obtain an enlarged image, comprising: an initializing means for setting an original image as an image to be enlarged; and a method based on a wavelet transform for the image to be enlarged. Is applied, a target image enlarging unit that generates an enlarged image having four times the number of pixels, and an enlarged image obtained by the target image enlarging unit is set as an enlargement target image, and the process returns to the target image enlarging unit. Means, an enlarged image presenting means for visually presenting the enlarged image obtained by the target image enlarging means, and an image adjusting means for performing enlarging processing or reducing processing on the enlarged image presented by the enlarged image presenting means And an enlarged image output unit that outputs an image obtained by the image adjustment unit. 7. The image enlargement unit according to claim 2, wherein the target image enlargement unit uses the target image instead of using the adjusted image. Image enlargement device. 8. A color image enlarging device for obtaining an enlarged image of Ln pixels × Lm pixels by enlarging an input color digital image of n pixels × m pixels, wherein a basic color component is selected from the color image components. Basic component selection means, conversion ratio derivation means for deriving a conversion ratio when deriving another color component from the basic color component selected by the basic component selection means, and converting the basic color component of the input original image to Ln / 2 pixels x Lm./2
A basic component image adjusting unit that adjusts to pixels; a basic image enlarging unit that generates an enlarged image by applying a method based on a wavelet transform to a basic color component of the adjusted image by the previous basic component image adjusting unit; A basic enlarged image adjusting unit that adjusts an enlarged image of the basic color component obtained by the image enlarging unit to a desired Ln pixel × Lm pixel; and a conversion ratio is derived to an enlarged image of the basic color component derived by the basic enlarged image adjusting unit. A missing component enlargement unit for estimating an enlarged image in another color component by applying the conversion ratio of the unit; an enlarged image of the basic color component derived by the basic enlarged image output unit; and an estimation by the insufficient component enlargement unit. Enlarged image output means for generating an enlarged image of a color original image by combining the remaining color components obtained, Location. 9. The color image enlarging device according to claim 8, wherein the image enlarging unit according to any one of claims 2 to 4 is used as the basic image enlarging unit. 10. An HL correction amount estimating means for correcting a difference image between an image of the reference HL component generating means and a sub-band component belonging to a vertically low frequency region of the image obtained by the level down means. The LH correction amount estimating means obtains, as a correction component, a difference image between the image of the reference LH component generating means and a subband component belonging to a horizontally low frequency region of the image obtained by the level down means. The HH correction amount estimating means obtains, as a correction component, a difference image between the image of the reference HH component generating means and a subband component belonging to a high frequency region of the image obtained by the level down means, and HL, LH, HH Each component estimating means performs a correction process by enlarging the three corrected component images into an image having four times the number of pixels by linear interpolation and adding the image to a corresponding edge image. Image enhancement apparatus according to claim 3, characterized. 11. An HL correction amount estimating means for correcting a difference image between an image of the base reference edge generating means and a subband component belonging to a vertically low frequency region of the image obtained by the level down means. The LH correction amount estimating means obtains, as a correction component, a difference image between the image of the reference edge generating means and a subband component belonging to a horizontally low frequency region of the image obtained by the level down means, The HH correction amount estimating means obtains, as a correction component, a difference image between an image of the reference edge generating means and a subband component belonging to a high frequency region of the image obtained by the level down means, and each of HL, LH, and HH The component estimating means performs a correction process by enlarging the three corrected component images into an image having four times the number of images by linear interpolation and adding the image to a corresponding edge image. Image enhancement apparatus according to claim 4, characterized.