JP2002165090A - Method and device for converting multiplex resolution and recording medium - Google Patents

Abstract

PROBLEM TO BE SOLVED: To obtain a band limited picture signal which is minutely divided in bands when an original picture signal is converted in multiplex resolution so as to obtain the band limited picture signal with more than one frequency bands. SOLUTION: An interpolating (a) being reduction in square checker is performed in the original picture signal S0 in a square pixel arrayal so as to obtain a reduced picture signal S1. An interpolating (b) being square checker magnification is performed in the signal S1 so as to obtain a magnified picture signal SB1. A difference between the original picture signal S0 and the magnified picture signal S1 is obtained and made to be the band limited picture signal B1. An interpolating (c) being the square checker reduction is performed in the reduced picture signal S1 so as to obtain a reduced picture signal S2. An interpolating (d) being the square checker magnification is performed in the picture signal S2 so as to obtain a magnified picture signal SB2. A difference between the reduced picture signal S1 and the magnified picture signal SB2 is obtained and adopted as a band limited picture signal B2. The processing is repeated to obtain a band limited picture signal Bk which expresses the picture of the more than one frequency bands of an original picture.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a method and an apparatus for multi-resolution conversion of an image signal to obtain a band-limited image signal representing an image for each of a plurality of frequency bands. The present invention relates to an image processing method and apparatus for performing processing, a multi-resolution conversion method, and a computer-readable recording medium storing a program for causing a computer to execute the image processing method.

[0002]

2. Description of the Related Art In various fields, an image signal representing an image is obtained, the image signal is subjected to appropriate image processing, and the processed image is reproduced and displayed. For example, in order to improve the diagnostic performance of a radiation image, a method of applying frequency emphasis processing such as blur mask processing to an image signal has been proposed by the present applicant (Japanese Patent Application Laid-Open No. Sho 55-163772).
etc). This frequency processing is processing for adding a signal obtained by subtracting a blur mask signal from an image signal representing an original image and multiplying the result by a degree of enhancement, thereby emphasizing a predetermined spatial frequency component in the image. is there.

As another method of performing frequency processing on an image signal, a Fourier transform, a wavelet transform,
By converting an image into a multi-resolution image by subband conversion or the like, an image signal representing the image is decomposed into signals in a plurality of frequency bands, and of the decomposed signals, a signal in a desired frequency band is emphasized. For example, a method of performing predetermined image processing such as the above has been proposed.

As a novel method for converting an image into a multi-resolution space, a Laplacian pyramid method has been proposed (for example, JP-A-5-244508 and JP-A-6-301766). In the Laplacian pyramid method, an image signal representing an original image is subjected to mask processing using a mask approximated by a Gaussian function, and then
By reducing the number of pixels by halving the image by sub-sampling, a reduced blurred image signal representing a reduced blurred image having a size of 1/4 of the original image is obtained, and a value of 0 is assigned to the sampled pixel of the reduced blurred image. The pixel of is returned to the image of the original size by interpolating, and the image is further subjected to mask processing by the above-described mask to obtain an enlarged blurred image signal representing an enlarged blurred image, and this enlarged blurred image is converted from the original image. Subtraction is performed to obtain a band-limited image signal representing a detailed image in a predetermined frequency band of the original image. By repeated for this process the resulting magnified blurred image, magnified blurred image signal representing a magnified blurred image size of 1/2 ^2N of the original image, and a 1/2 ^2N size detail image of the original image This is to create N band-limited image signals to be represented. In this processing, since sampling is performed on an image subjected to mask processing using a mask approximated by a Gaussian function, a Gaussian filter is actually used, but when filtering processing is performed using a Laplacian filter A detailed image similar to the above is obtained.

[0005] By such processing, a band-limited image signal representing a detail image for each frequency band having a size of 1 / ^22N is obtained in order from an image of the original image size. This detail image is called a Laplacian pyramid. You. Note that an enlarged blurred image having a size of 1/2 ^2N is obtained in order from the image having the original image size, and this enlarged blurred image is called a Gaussian pyramid.

The Laplacian pyramid is described in Burt PJ, “Fast Filter Transforms for Image P.
rocessing ”, Computer Graphics and Image Processi
ng 16: 20-51, 1981; Crowley JL, Stern R.
M., “Fast Computation of the Difference of Low Pass
Transform ”IEEETrans.on Pattern Analysis and Mach
ine Intelligence, Volume 6, Issue 2, March 1984, Mallat
SG, “A Theory for Multiresolution Signal Decompo
sition ； The Wavelet Representation ”IEEE Trans.on
Pattern Analysis and Machine Intelligence, 11
Vol. 7, No. 7, July 1989; Ebrahimi T., Kunt M., "Image
compression by Gabor Expansion ”, OpticalEngineeri
ng, Vol. 30, No. 7, pp. 873-880, July 1991, and Pie
ter Vuylsteke, Emile Schoeters, “Multiscale Image
Contrast Amplification ”SPIE Vol.2167 Image Proc
The details are described in essing (1994), pp. 551 to 560.

The band-limited image signals of all the frequency bands of the Laplacian pyramid thus obtained are subjected to an emphasis process using, for example, a nonlinear function for suppressing the obtained signal value as the signal value increases. A method of obtaining a processed image signal by inversely converting the band-limited image signal of each frequency band to which the emphasis processing has been applied is disclosed in Japanese Patent Laid-Open No. 5-24 / 1993.
No. 4508 and JP-A-6-301766. Since the detail image of each frequency band is enhanced by such enhancement processing, the image is substantially an image obtained by performing the blur mask processing using a plurality of size masks in the above-described blur mask processing. Further, since a non-linear function that suppresses the signal value as the signal value increases is used, it is possible to prevent a sharp edge included in the original image from being excessively emphasized.

In addition, the above emphasis processing is not performed on band-limited image signals of all frequency bands. An image processing method in which a frequency band to be processed is selected according to a shooting menu has also been proposed (Japanese Patent Laid-Open No. 9-161061).

By the way, among the detailed images for each of a plurality of frequency bands obtained by converting an image into a Laplacian pyramid, a detailed image in a relatively high frequency band represents a relatively fine portion to be observed. Therefore, the signal value at a certain pixel indicates a difference in contrast between the pixel and a relatively close pixel. On the other hand, a detail image in a relatively low frequency band indicates a gradual change in the background of the original image, and therefore, a signal value at a certain pixel indicates a difference in contrast with a pixel relatively far from the pixel. . Therefore, the pixel value in the detail image in the relatively low frequency band has a larger value than the pixel value in the detail image in the high frequency band.

In view of this point, among the images of a plurality of frequency bands converted into the multi-resolution space, the pixel value of each pixel is smaller than a predetermined threshold value for a detailed image of a predetermined frequency band to be enhanced. Is larger than the other pixels, a method has been proposed in which the emphasis degree is made smaller than the emphasis degree of the other pixels to perform the emphasis processing (Japanese Patent Application Laid-Open No. 9-44657). According to this method, the degree of enhancement is reduced in a detail image in a low frequency band having a relatively large pixel value, and the degree of enhancement is increased in a detail image in a high frequency band having a relatively small pixel value. Even when the density around the part to be observed is relatively high and the part to be observed is difficult to see, the degree of emphasis around the part to be observed is reduced and this part is emphasized. In the completed image, the target portion can be made easier to observe.

[0011]

In the method of converting an image into a multi-resolution space using the Laplacian pyramid, a band-limited image signal representing a detail image having a size of 1 / ^22N of the original image is obtained. The frequency band of each band-limited image signal is の of the Nyquist frequency of the original image signal
^It represents a frequency band at ^N intervals. Therefore, it is possible to perform processing at a frequency interval of 1/2 ^N of the Nyquist frequency. On the other hand, half of the Nyquist frequency
It is desired to obtain a band-limited image signal that is divided into frequency bands finer than the ^N frequency bands.

The present invention has been made in view of the above circumstances, and a first object of the present invention is to provide a multi-resolution conversion method and apparatus capable of multi-resolution conversion of an image into a finer frequency band.
It is the purpose of.

It is a second object of the present invention to provide an image processing method and apparatus capable of performing image processing in a finer frequency band.

A further object of the present invention is to provide a computer-readable recording medium in which a program for causing a computer to execute the above-described multi-resolution conversion method and image processing method is recorded.

[0015]

According to a first multi-resolution conversion method of the present invention, a multi-resolution conversion process is performed on an original image signal having a square pixel array to obtain an original image represented by the original image signal. A multi-resolution conversion method for obtaining a band-limited image signal representing an image of each of a plurality of frequency bands, wherein a first reduced image signal is obtained by performing square check processing on the original image signal. A second step of obtaining a first enlarged image signal by performing a checkerboard square enlargement process on the first reduced image signal, and calculating a difference value between the original image signal and the first enlarged image signal. A third step of obtaining a first band-limited image signal,
A fourth step of obtaining a second reduced image signal by performing a checkerboard square reduction process on the first reduced image signal, and performing a square checkerboard enlargement process on the second reduced image signal. A fifth step of obtaining a second enlarged image signal, and obtaining a difference value between the first reduced image signal and the second enlarged image signal, so that the frequency is one step lower than the first band-limited image signal. A sixth step of obtaining a second band-limited image signal of a band, and the first to third steps for the second reduced image signal until a band-limited image signal of a desired frequency band is obtained. It has a seventh step of repeating the first to sixth steps.

In a second multi-resolution conversion method according to the present invention, a multi-resolution conversion process is performed on an original image signal having a checkered pixel array, and a plurality of frequency bands of an original image represented by the original image signal are processed. A first step of obtaining a first reduced image signal by performing a square checkerboard reduction process on the original image signal, wherein the first step comprises: A second step of obtaining a first enlarged image signal by performing a square checkerboard enlargement process on the reduced image signal, and obtaining a difference value between the original image signal and the first enlarged image signal, A third step of obtaining a band-limited image signal, a fourth step of performing a square checkerboard reduction process on the first reduced image signal to obtain a second reduced image signal, and the second reduced image signal. Against the city A fifth step of obtaining a second enlarged image signal by performing a square enlargement process; and obtaining a difference value between the first reduced image signal and the second enlarged image signal, thereby obtaining the first band limitation. A sixth step of obtaining a second band-limited image signal of a lower frequency band by one stage than the image signal, and the second step of obtaining the second band-limited image signal of a desired frequency band until the second reduced image signal is obtained. The method has a seventh step of repeating the first to third steps or the first to sixth steps.

The "square pixel arrangement" means that the pixels of the image represented by the image signal are arranged such that the nearest pixels are arranged at equal vertical and horizontal intervals as shown in FIG.

The "checkered pixel array" means that the pixels of the image represented by the image signal are arranged at equal intervals in the diagonal direction, as shown in FIG. An arrangement in which each pixel of a pixel row adjacent to the pixel row is positioned therebetween.

"Square checkerboard reduction processing" refers to filtering processing based on the pixel values of adjacent pixels in the vertical, horizontal, and vertical directions with respect to an image signal of a square pixel array, by alternately thinning out each pixel in an adjacent pixel row. This refers to a process of changing the pixel arrangement from a square shape to a checkered shape while reducing the number of pixels by performing. By this square checkerboard reduction process, the processed image signal (first reduced image signal) has a checkerboard pixel array, and the number of pixels is approximately の of the image signal of the original square pixel array.
Becomes

The "checkerboard square enlargement process" is a process of interpolating pixels between adjacent pixels on an image signal of a checkerboard pixel array, thereby increasing the number of pixels and changing the pixel array from a checkerboard pattern. This refers to the process of changing to a square shape. By this checkerboard square enlargement process, the processed image signal (first enlarged image signal) has a square pixel array, and the number of pixels is approximately twice as large as the image signal of the checkerboard pixel array.

In the first multi-resolution conversion method according to the present invention, the first enlarged image signal obtained by subjecting the original image signal to square checkerboard reduction processing and checkerboard square enlargement processing is the same as the original image signal. Are image signals of a square pixel array and have the same number of pixels as the original image signal. On the other hand, in the second multi-resolution conversion method according to the present invention, the first enlarged image signal obtained by performing the checkerboard square reduction process and the square checkerboard enlargement process on the original image signal is similar to the original image signal. It is an image signal with a checkered pixel array and has the same number of pixels as the original image signal. Therefore, "determining a difference value between the original image signal and the first enlarged image signal" means performing a subtraction process between corresponding pixels in the original image signal and the first enlarged image signal. As a result, it is possible to obtain a band-limited image signal consisting of only the frequency components of the highest frequency band in the original image signal.

The "checkerboard square reduction process" is a process of alternately thinning out each pixel of an adjacent pixel row from an image signal of a checkerboard pixel array, or a filtering process based on pixel values of diagonally adjacent pixels. By changing the pixel arrangement from a checkered pattern to a square pattern while reducing the pixel value. By this checkerboard square reduction process, the processed image signal (second reduced image signal) becomes a square array,
The number of pixels is approximately の of the original image signal of the checkerboard arrangement.

The "square checkerboard enlargement process" is a process of interpolating a pixel at an intersection of diagonal lines in a case where a rectangular image having four pixels adjacent to each other as vertices is considered for an image signal of a square pixel array. Is performed to change the pixel arrangement from a square shape to a checkered shape while increasing the number of pixels. By this square checkerboard enlargement processing, the processed image signal (second enlarged image signal) becomes a checkerboard pixel array,
The number of pixels is approximately double as compared with the image signal of the square pixel array.

In the first multi-resolution conversion method according to the present invention, the first reduced image signal is subjected to a checkerboard square reduction process and a square checkerboard enlargement process.
Is an image signal having a checkerboard pixel array like the first reduced image signal, and has the same number of pixels as the first reduced image signal. On the other hand, in the second multi-resolution conversion method according to the present invention, the second enlarged image signal obtained by performing the square checkerboard reduction process and the checkerboard square enlargement process on the first reduced image signal is the first enlarged image signal. It is an image signal of a square pixel array like the reduced image signal,
And it has the same number of pixels as the first reduced image signal. Therefore, “calculating the difference value between the first reduced image signal and the second enlarged image signal” means that the subtraction process is performed between the corresponding pixels in the first reduced image signal and the second enlarged image signal. To do. Accordingly, it is possible to obtain a second band-limited image signal having only one frequency component in a frequency band corresponding to the first reduced image signal and having a lower frequency band by one step than the first band-limited image signal.

"Repeat the first to third steps or the first to sixth steps" means that the band-limited image signal of the desired frequency band is in a checkered pixel array (in the second multi-resolution conversion method, If the image signal is a square pixel array, the first to third steps are repeated, and the band-limited image signal of the desired frequency band is changed to a square pixel array (in the second multi-resolution conversion method, a checkered pattern). In the case of an image signal of a (shape pixel array), the first to sixth steps are repeated.

"Repeat" means that the second reduced image signal is regarded as the original image signal, and the square reduced check processing of the first step is performed on the second reduced image signal (in the second multi-resolution conversion method, the checkerboard is checked). Square reduction processing), the checkered square enlargement processing (square checkered enlargement processing in the second multi-resolution conversion method) of the second step for the reduced image signal obtained in the first step, and the second step. The process of the third step for obtaining a difference value between the enlarged image signal and the second reduced image signal, and the fourth step of the checkerboard square reduction process on the reduced image signal obtained in the second step (the second step) In the multi-resolution conversion method, a square checkerboard reduction process), and in a fifth step, a square checkerboard enlargement process (second multi-resolution) for the reduced image signal obtained in the fourth step. In the conversion method, it means that the processing of the sixth step for obtaining a difference value between the reduced image signal obtained in the first step and the enlarged image signal obtained in the fifth step is performed. . Further, if necessary, the processing of the first to third or first to sixth steps may be further performed after the processing of the first to sixth steps is performed for the second time. As a result, a band-limited image signal from the highest frequency band to the desired frequency band is obtained stepwise.

The signals obtained in the first and second multi-resolution conversion methods according to the present invention correspond to the band-limited image signal for each of the plurality of frequency bands and the second reduced image signal corresponding to the frequency band. A reduced image signal (when the first to sixth steps are performed) or a reduced image signal corresponding to the first reduced image signal (when the first to third steps are performed).

[0028] The image processing method according to the present invention is a method for processing a plurality of frequency bands obtained by the first or second multi-resolution conversion method according to the present invention. The band-limited image signal is subjected to predetermined image processing such as enhancement processing and noise suppression processing.

"Band-limited image signal of predetermined frequency band"
Means that a band-limited image signal representing an image for each of a plurality of frequency bands up to a desired frequency band can be obtained by the first and second multi-resolution conversion methods according to the present invention. Of which should be subjected to image processing
Or a band-limited image signal of a higher frequency band.

In the image processing method according to the present invention, the band-limited image signal subjected to the predetermined image processing and the other band-limited image signal are processed by performing inverse multi-resolution conversion processing. It is preferable to obtain an image signal.

The "inverse multi-resolution conversion process" is a process in which the processes performed in the first to sixth steps are performed in reverse order from the sixth step on a band-limited image signal of a desired frequency band. Means obtaining a processed image signal representing a processed image of the same size as the original image.
More specifically, assuming that the processes of the first to sixth steps are performed only once and the band-limited image signal is not subjected to image processing, the square checkerboard enlargement process is performed on the second reduced image signal. (In the second multi-resolution conversion method, a square checker enlargement process is performed) to obtain a second enlarged image signal, and the sum of the second enlarged image signal and the band-limited image signal of the desired frequency band is calculated. The first reduced image signal is obtained to obtain the first reduced image signal, and the first reduced image signal is square-checked square enlargement processing (square checkered enlargement processing in the second multi-resolution conversion method)
Is performed, a first enlarged image signal is obtained, and a sum of the first enlarged image signal and the first band-limited image signal is obtained to obtain an original image signal.

A first multi-resolution conversion apparatus according to the present invention performs multi-resolution conversion processing on an original image signal having a square pixel array, and converts the original image signal represented by the original image signal into a plurality of frequency bands. A first means for obtaining a first reduced image signal by subjecting said original image signal to a square checkerboard reduction process, said first means for obtaining a band-limited image signal representing said image; A second unit for obtaining a first enlarged image signal by performing a checkerboard enlargement process on the reduced image signal; and obtaining a first difference value between the original image signal and the first enlarged image signal. A third means for obtaining a band-limited image signal, a fourth means for obtaining a second reduced image signal by performing a checkerboard reduction process on the first reduced image signal, and the second reduced image signal For the square checkerboard enlargement process Fifth means for obtaining a second enlarged image signals by Succoth, the first reduced image signal and the second
By calculating a difference value between the first band-limited image signal and the second band-limited image signal,
A sixth means for obtaining a band-limited image signal of the above, until the band-limited image signal of a desired frequency band is obtained, the processing in the first to third means on the second reduced image signal or the A multi-resolution conversion means for repeating the processing in the first to sixth means is provided.

A second multi-resolution conversion device according to the present invention performs multi-resolution conversion processing on an original image signal having a checkerboard pixel array, and converts the original image signal represented by the original image signal into a plurality of frequency bands. A first means for obtaining a first reduced image signal by performing a checkerboard square reduction process on the original image signal, wherein the first means for obtaining a band-limited image signal representing the image A second unit that obtains a first enlarged image signal by performing a square checkerboard enlargement process on the reduced image signal; and obtains a first value by calculating a difference value between the original image signal and the first enlarged image signal. A third means for obtaining a band-limited image signal, a fourth means for performing a square checkerboard reduction process on the first reduced image signal to obtain a second reduced image signal, and the second reduced image signal. Ichimatsu Masakata expansion processing Fifth means for obtaining a second enlarged image signals by Succoth, the first reduced image signal and the second
By calculating a difference value between the first band-limited image signal and the second band-limited image signal,
A sixth means for obtaining a band-limited image signal of the above, until the band-limited image signal of a desired frequency band is obtained, the processing in the first to third means on the second reduced image signal or the A multi-resolution conversion means for repeating the processing in the first to sixth means is provided.

The image processing apparatus according to the present invention comprises a band-limited image signal representing an image for each of the plurality of frequency bands obtained by the first or second multi-resolution conversion apparatus according to the present invention. An image processing means for performing predetermined image processing on the band-limited image signal is provided.

In the image processing apparatus according to the present invention, it is preferable that the predetermined image processing is enhancement processing or noise suppression processing.

Further, in the image processing apparatus according to the present invention, the band-limited image signal subjected to the predetermined image processing and the other band-limited image signal are processed by performing inverse multi-resolution conversion processing. It is preferable to further include an inverse multi-resolution conversion unit for obtaining an image signal.

The first and second multi-resolution conversion methods and the image processing method according to the present invention may be provided by being recorded on a computer-readable recording medium as a program for causing a computer to execute the method.

[0038]

According to the first multi-resolution conversion method and apparatus of the present invention, a square checkerboard reduction process is performed on an original image signal having a square pixel array to obtain a first reduced image signal having a checkerboard pixel array. And applying a checkerboard square enlargement process to the first reduced image signal to obtain a first enlarged image signal of a square pixel array, and calculating a difference value between the original image signal and the first enlarged image signal. As a result, a first band-limited image signal is obtained. This first band limited image signal is an image signal of a checkerboard pixel array. Further, the first reduced image signal is subjected to a checkerboard square reduction process to obtain a second reduced image signal having a square pixel array, and the second reduced image signal is subjected to a square checkerboard enlargement process to perform a checkerboard checkerboard process. A second enlarged image signal having a pixel array is obtained, and a difference value between the first reduced image signal and the second enlarged image signal is obtained. 2 band-limited image signals are obtained.
This second band limited image signal is an image signal of a square pixel array. Then, by repeating the processing of the first to third steps or the processing of the first to sixth steps for the second reduced image signal, the band-limited image signal of the square pixel array and the band-limited image signal of the checkered pixel array Image signals are obtained alternately in steps up to a desired frequency band.

On the other hand, according to the second multi-resolution conversion method and apparatus according to the present invention, a checkerboard square reduction process is performed on an original image signal having a checkerboard array to generate a first reduced image signal having a square pixel array. Then, the first reduced image signal is subjected to square checkerboard enlargement processing to obtain a first enlarged image signal having a checkered pixel array, and a difference value between the original image signal and the first enlarged image signal is obtained. , A first band-limited image signal is obtained. This first band limited image signal is an image signal of a checkerboard pixel array. Further, the first reduced image signal is subjected to square checkerboard reduction processing to obtain a second reduced image signal having a checkered pixel array, and the second reduced image signal is subjected to checkerboard square enlargement processing to be squared. A second enlarged image signal having a pixel array is obtained, and a difference value between the first reduced image signal and the second enlarged image signal is obtained. 2 band-limited image signals are obtained.
This second band limited image signal is an image signal of a square pixel array. Then, by repeating the processing of the first to third steps or the processing of the first to sixth steps for the second reduced image signal, the band-limited image signal of the checkerboard pixel array and the band-limited of the square pixel array are Image signals are obtained alternately in steps up to a desired frequency band.

Here, the band-limited image signal obtained by the above-described Laplacian pyramid method represents a detailed image in a frequency band of 1/2 ^N to 1/2 ^{N + 1} of the Nyquist frequency of the original image signal. ing. On the other hand, the band-limited image signal obtained by the present invention is ナ イ^N to ^{k k} and 〜 ^k to ^{N N +1} of the Nyquist frequency of the original image signal.
It represents a detailed image in a frequency band of (N <k <N + 1). For example, while the second band-limited image signal represents a detail image having a size of 1/2 ² of the original image, the first band-limited image signal is composed of the original image and 1/1/2 of the original image.
It has to represent the detail image of the intermediate size of 2 ² size of detail images. Therefore, if the Nyquist frequency of the original image signal is fs / 2, the first band-limited image signal is fs
s / 2 to fs / k, and the second band-limited image signal is fs / k
Fs / 4 (2 <k <4). Therefore, according to the present invention, it is possible to obtain a band-limited image signal obtained by dividing the original image signal into smaller frequency bands as compared with the conventional Laplacian pyramid method.

Further, by performing image processing such as enhancement processing and noise suppression processing on a band-limited image signal of a predetermined frequency band among band-limited image signals representing images of a plurality of frequency bands, Since image processing can be performed at fine frequency band intervals, image quality can be controlled in more detail. Therefore, a processed image signal obtained by performing inverse multi-resolution conversion of the processed band-limited image signal and the unprocessed band-limited image signal can be used to reproduce a higher-quality processed image. It can be.

In the Laplacian pyramid method, the original image signal having a square pixel array is divided into bands to obtain a band-limited image signal having a square array. When it is desired to suppress sharp edges, edges can be favorably suppressed in the vertical and horizontal directions, that is, in the direction in which pixels are arranged in a square shape.
However, since the band-limited image signal is not band-divided in the oblique direction, the edge cannot be satisfactorily suppressed in the oblique direction. On the other hand, the band-limited image signal obtained by the present invention is obtained by using a band-limited image signal of a checkerboard pixel array that is band-divided in an oblique direction because the original image is band-divided into a square and a checkerboard. ,
Edges can also be satisfactorily suppressed in oblique directions. Therefore, it is possible to prevent the edge from being excessively emphasized regardless of the direction of the edge.

[0043]

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

FIG. 1 is a schematic block diagram showing the configuration of an image processing system provided with a multi-resolution conversion device and an image processing device according to the first embodiment of the present invention. As shown in FIG. 1, the image processing system performs a multi-resolution conversion process on an original image signal S0 representing an original image and converts the original image signal S0 into an image (detailed image) for each of a plurality of frequency bands of the original image.
Resolution converting means 1 for obtaining a band-limited image signal representing
And an emphasis processing unit 2 that performs an emphasis process as described later on the band-limited image signal obtained by the multi-resolution conversion unit 1 to obtain a processed band-limited image signal, and demultiplexes the processed band-limited image signal. Inverse multi-resolution conversion means 3 for obtaining a processed image signal Sproc by resolution conversion, and a processed image signal Sproc obtained by the inverse multi-resolution conversion means 3
And an image output unit 4 including a printer, a monitor, and the like for reproducing the image as a visible image, and a table storage unit 5 for storing a table T representing the degree of enhancement of the enhancement processing performed by the enhancement processing unit 2.

FIG. 2 is a diagram for explaining the processing performed in the multi-resolution conversion means 1, the enhancement processing means 2, and the inverse multi-resolution conversion means 3. Multi-resolution conversion means 1
Converts the original image signal S0 into a band-limited image signal representing a detailed image for each of a plurality of frequency bands of the original image as described below. Here, the original image represented by the original image signal S0 is one in which pixels are arranged in a square shape as shown in FIG. Note that, in FIG. 3 and subsequent figures, the position of a circle represents the position of a pixel.

First, an interpolation process a, which is a square checkerboard reduction process, is performed on the original image signal S0. Note that interpolation processing a
Means corresponds to the first means in the first multi-resolution conversion device according to the present invention. This interpolation processing a is performed as shown in FIG.
Is performed by filtering each pixel of the original image signal S0 every other pixel by the filter shown in FIG. This filtering process is performed by shifting pixels in adjacent pixel columns one by one. This allows
The filtering process is performed only on the pixel at the position ○ shown in FIG. 5, and a reduced image signal S1 representing a reduced image in which the pixels are arranged in a checkered pattern as shown in FIG. 6 is obtained. Further, instead of the filtering process, a process of thinning out pixels every other pixel may be simply performed.

The reduced image signal S1 has a pixel arrangement in a checkerboard pattern compared to the original image signal S0, and the number of pixels is approximately ２. Note that “approximately ２” means that the interpolation process “a” is a process for reducing the number of pixels to １ ／. This is because there is more than one pixel, and it is not completely halved.

Next, the reduced image signal S1 is subjected to an interpolation process b which is a checkerboard square enlargement process. The means for performing the interpolation processing b corresponds to the second means in the first multi-resolution conversion device according to the present invention. In this processing, the pixel value of the position of the cross mark in FIG. 5 of the reduced image signal S1 is obtained, and the enlarged image signal SB representing the enlarged image having the same number of pixels as the original image is obtained.
This is the process for obtaining the number 1. Specifically, for the position of the mark “○” in FIG. 6, the pixel value is used as it is, and for the position of the mark “×” in FIG. 5, the filtering process is performed by the filter shown in FIG. To determine the pixel value. Thus, as shown in FIG. 8, an enlarged image signal SB1 representing an enlarged image in which pixels are arranged in a square shape is obtained. In FIG. 8, ８ indicates a pixel position obtained by the interpolation processing b.

Then, by subtracting the enlarged image signal SB1 from the original image signal S0, a band-limited image signal B1 of the highest frequency band of the original image signal S0 is obtained. here,
Since the images represented by the original image signal S0 and the enlarged image signal SB1 have the same number of pixels and both have a square pixel arrangement, the difference between the corresponding pixels of the original image signal S0 and the enlarged image signal SB1 is determined. By obtaining the value, a band-limited image signal B1 representing a detailed image having a square pixel array is obtained.

Further, the reduced image signal S1 obtained by the above-mentioned interpolation processing a is subjected to an interpolation processing c which is a checkerboard square reduction processing. The means for performing the interpolation processing c corresponds to the fourth means in the first multi-resolution conversion device according to the present invention. This interpolation processing c is performed by filtering each pixel row of the reduced image signal S1 every other row by the filter shown in FIG. As a result, FIG.
The filtering process is performed only on the pixel at the position ○ indicated by 0, and a reduced image signal S2 representing a reduced image in which pixels are arranged in a square as shown in FIG. 11 is obtained. Note that a process of simply thinning out pixels may be performed instead of the filtering process.

The reduced image signal S2 is a reduced image signal S1.
And the number of pixels is approximately ２.
It has become. Further, the reduced image signal S2 has half the number of pixels in the vertical and horizontal directions, that is, the number of pixels is ４ compared to the original image signal S0. Therefore, the reduced image signal S2
The size of the image represented by is 1/2 ² of the original image.

Next, an interpolation process d, which is a square checkerboard enlargement process, is performed on the reduced image signal S2. The means for performing the interpolation processing d corresponds to the fifth means in the first multi-resolution conversion device according to the present invention. This process is a process of obtaining a pixel value of the reduced image signal S2 at the position of the cross mark in FIG. 10 to obtain an enlarged image signal SB2 representing an enlarged image having the same number of pixels as the image represented by the reduced image signal S1.
More specifically, the pixel values as they are at the positions indicated by the circles in FIG. 10 are used, and the positions at the crosses shown in FIG.
A pixel value is obtained by performing a filtering process using the filter shown in 2 (a) or (b). Thus, as shown in FIG. 13, an enlarged image signal SB2 representing an enlarged image in which pixels are arranged in a checkered pattern is obtained. Note that, in FIG. 13, ◎ indicates a pixel position obtained by the interpolation processing d.

Then, by subtracting the enlarged image signal SB2 from the reduced image signal S1, the band-limited image signal B1 is obtained.
Thus, a band-limited image signal B2 in a lower frequency band by one step than that is obtained. Here, since the images represented by the reduced image signal S1 and the enlarged image signal SB2 both have a checkered pixel arrangement, the reduced image signal S1 and the enlarged image signal SB
By calculating the difference value between the two corresponding pixels,
A band-limited image signal B2 representing a detailed image having a checkered pixel arrangement is obtained.

The reduced image signal S2 is subjected to interpolation processes a to d in the same manner as described above, and the reduced image signal Sk (k = 1 to n) obtained by the interpolation process and the enlarged image signal SBk (k = 1 to n), the difference value is obtained.
Band-limited image signals Bk (k = 1 to
n) is obtained. Here, when k is an odd number, the reduced image represented by the reduced image signal Sk has a checkered pixel array, and the enlarged image represented by the enlarged image signal SBk and the detailed image represented by the band-limited image signal Bk are: It becomes a square pixel array. On the other hand, when k is an even number, the reduced image represented by the reduced image signal Sk has a square pixel array, and the enlarged image represented by the enlarged image signal SBk and the detailed image represented by the band-limited image signal Bk are checkered. Pixel arrangement. When the band-limited image signal Bn of the lowest frequency band is obtained by using the reduced image signal Sn obtained by performing the interpolation processing c, the detailed image represented by the band-limited image signal Bn is a checkered pixel. It becomes an array. On the other hand, the reduced image signal S obtained by performing the band-limited image signal Bn of the lowest frequency band up to the interpolation process a
When using n, the detailed image represented by the band-limited image signal Bn has a square pixel array.

The emphasis processing means 2 performs an emphasis processing on the band-limited image signal Bk based on the table T stored in the table storage means 5. FIG. 14 is a diagram showing an example of the table T stored in the table storage means 5. As shown in FIG. 14, the table T suppresses the signal when the absolute value of the signal value of the band-limited image signal Bk is larger than the threshold α, and emphasizes the signal when the absolute value of the signal value is smaller than the threshold α. Is a non-linear function. Here, FIG.
In FIG. 4, M indicates a range of possible values of the pixel. For example, when the range of possible values of a pixel is 10 bits, M = 1
023. In such a case, the threshold α is 256
About. Then, based on this table T, the band-limited image signal Bk is subjected to enhancement processing to obtain a processed band-limited image signal Bk ′.

By such an emphasizing process, the edge component of each frequency band included in the original image is emphasized, but a sharp edge having a large signal value in the band-limited image signal Bk is excessively emphasized. Is prevented.

It is to be noted that the emphasis processing is not performed on the band-limited image signal Bk of all the frequency bands, and the image pickup when the original image signal S0 is obtained as described in, for example, JP-A-9-161061. The band-limited image signal Bk of the frequency band to be emphasized may be selected based on a shooting menu such as a part and shooting conditions.

Further, the obtained image signal is transferred to a hard disk, C
There is a case where the data is temporarily stored in a recording medium such as a DR and an FD. In this case, the stored image signals are the band-limited image signal Bk of all the frequency bands and the reduced image signal Sn of the lowest frequency band.

In the inverse multi-resolution conversion means 3, inverse multi-resolution conversion is performed on the processed band-limited image signal Bk 'as follows, and the processed image signal Sproc
Is obtained. Here, it is assumed that the multi-resolution conversion means 1 has performed up to the interpolation processing c to obtain the reduced image signal Sn of the lowest frequency band of the square pixel array. In this case, the detailed image represented by the band-limited image signal Bn in the lowest frequency band has pixels arranged in a checkered pattern. First, the reduced image signal Sn of the lowest frequency band is subjected to an interpolation process d, which is a square checkerboard enlargement process,
An enlarged image signal SBn representing an enlarged image in which pixels are arranged in a checkered pattern is obtained. Note that this enlarged image signal SBn is
Enlarged image signal S obtained by multi-resolution conversion means 1
It is the same as Bn. Then, an addition signal San-1 is obtained by adding the processed band limited image signal Bn ′ of the lowest frequency band and the enlarged image signal SBn.
Here, since the images represented by the enlarged image signal SBn and the processed band-limited image signal Bn 'both have a checkered pixel arrangement, the correspondence between the enlarged image signal SBn and the processed band-limited image signal Bn' By obtaining an added value of the pixels to be obtained, an added signal San-1 representing an image having a checkered pixel array is obtained.

Next, an interpolation process b, which is a checkerboard square enlargement process, is performed on the addition signal San-1, and an enlarged image signal SBn-1 'representing an enlarged image in which pixels are arranged in a square shape.
Is obtained. The enlarged image signal SBn-1 'is added to the processed band-limited image signal Bn-1' in the frequency band one step higher than the lowest frequency band, and an addition signal San-2 is obtained. Here, since the images represented by the enlarged image signal SBn-1 'and the processed band-limited image signal Bn-1' both have a square pixel array, the enlarged image signal SBn-1 'and the processed band By obtaining the sum of the pixels corresponding to each other in the restricted image signal Bn-1 ', an sum signal San-2 representing an image having a checkerboard pixel arrangement is obtained.

Hereinafter, the interpolation processing b for the addition signal Sak
(K: interpolation process d for odd and even numbers) and addition of the expanded image signal SBk 'obtained by the interpolation processes b and d and the processed band-limited image signal Bk' are performed on the expanded image signal SB1 'of the highest frequency band. This is repeated until addition with the processed band-limited image signal B1 'of the highest frequency band is performed, to obtain an addition signal Sa0 of the highest frequency band. The sum signal Sa0 of the highest frequency band is the processed image signal Sproc.
Becomes

FIG. 15 is a diagram schematically showing the multi-resolution conversion processing, enhancement processing, and inverse multi-resolution conversion processing performed in the first embodiment. In FIG. 15, it is assumed that a band-limited image signal B4 from the highest frequency band to the four-step low-frequency band is obtained. A square pixel image signal is square, and a checkered pixel array image signal is square. Are represented by diamonds rotated 45 degrees. As shown in FIG. 15, when performing the multi-resolution conversion, the reduced image signals S1 and S3 of the checkered pixel array and the reduced image signals S2 and S3 of the square pixel array are obtained by the interpolation processes a and c.
4 are obtained alternately, and the enlarged image signals SB1 and SB3 of the square pixel array and the enlarged image signals SB2 and SB4 of the checkerboard pixel array are obtained alternately by the interpolation processes b and d. The difference values between the enlarged image signal, the original image signal, and the reduced image signal having the same pixel array are obtained for each frequency band, so that the band-limited image signals B1 and B3 of the square pixel array and the band of the checker pixel array are obtained. It can be seen that the restricted image signals B2, B4 are obtained alternately stepwise.

On the other hand, when performing the inverse multi-resolution conversion, the enlarged image signal SB of the checkered pixel array is
4, SB2 'and an enlarged image signal SB of a square pixel array
3 'and SB1' are obtained alternately, and the sum of the enlarged image signal and the processed band-limited image signal corresponding to each other, that is, having the same number of pixels and the same pixel arrangement, is obtained for each frequency band. Pixel array addition signals Sa3, S
a1 and addition signals Sa2 and Sa0 of the square pixel array
It can be seen that (Sproc) is obtained alternately for each frequency band.

Next, the operation of the first embodiment will be described. FIG. 16 is a flowchart showing the operation of the first embodiment. First, in the multi-resolution conversion means 2,
A multi-resolution conversion process is performed on the original image signal S0 to obtain a band-limited image signal Bk representing a detailed image for each of a plurality of frequency bands (step S1). FIG. 17 is a flowchart showing the operation of the multi-resolution conversion process. Since the original image signal S0 has a square pixel array, first, an interpolation process a which is a square checkerboard reduction process is performed on the original image signal S0.
Is performed to obtain a reduced image signal S1 of a checkerboard pixel array having approximately half the number of pixels as compared to the original image signal S0 (step S11). Next, the reduced image signal S1 is subjected to an interpolation process b, which is a checkerboard square enlargement process, to obtain an enlarged image signal SB1 having a square pixel array having the same number of pixels as the original image signal S0 (step S12). Then, by subtracting the enlarged image signal SB1 from the original image signal S0, a band-limited image signal B1 of the highest frequency band is obtained (step S13). Next, it is determined whether or not a band-limited image signal up to a desired frequency band has been obtained (step S14), and the process ends when step S14 is affirmed.

On the other hand, if step S14 is denied, the reduced image signal S1 is subjected to an interpolation process c, which is a checkerboard square reduction process, and compared with the reduced image signal S1 by approximately 1/2.
The reduced image signal S2 of the square pixel array having the number of pixels is obtained (step S15). Next, the reduced image signal S
2 is subjected to an interpolation process d, which is a square checkerboard enlargement process, to obtain an enlarged image signal SB2 having a checkerboard pixel array having the same number of pixels as the reduced image signal S1 (step S1).
6). Then, from the reduced image signal S1 to the enlarged image signal SB
By subtracting 2, the band-limited image signal B2 in the lower frequency band by one step than the highest frequency band is obtained (step S17). Next, it is determined whether or not a band-limited image signal up to the desired frequency band has been obtained (step S18), and the process ends when step S14 is affirmed.

On the other hand, if step S17 is negative, the process returns to step S11, and the processes from step S11 to step S18 are repeated for the reduced image signal S2. Thereby, the band-limited image signal B up to the desired frequency band
k is obtained.

Returning to FIG. 16, when the band-limited image signal Bk up to the desired frequency band is obtained, the emphasis processing means 2
, An emphasis process is performed on the band-limited image signal Bk of each frequency band, and the processed band-limited image signal Bk ′ is processed.
Is obtained (step S2).

When the processed band-limited image signal Bk 'is obtained, the inverse multi-resolution conversion unit 3 performs inverse multi-resolution conversion on the processed band-limited image signal Bk'. FIG. 18 is a flowchart showing the operation of the inverse multi-resolution conversion process. In the first embodiment, it is assumed that the band-limited image signal Bk has been obtained by performing the processing up to the interpolation c in the multi-resolution conversion means 1. Since the reduced image signal Sn in the minimum frequency band has a square pixel array, the reduced image signal Sn is first subjected to an interpolation process d, which is a square checkerboard enlargement process, to obtain a processed band-limited image in the lowest frequency band. An enlarged image signal Sn having a checkered pixel array having the same number of pixels as the signal Bn 'is obtained (step S21). Next, the processed band limited image signal Bn ′ of the lowest frequency band and the enlarged image signal Sn are added to obtain an addition signal San (Step S22). And
It is determined whether an addition signal of the highest frequency band has been obtained (step S23), and the process ends when step S23 is affirmed.

On the other hand, if step S23 is denied, the addition signal San is subjected to an interpolation process b, which is a square expansion process of the Ichimatsu, and a processed band-limited image of a frequency band one step higher than the lowest frequency band. Enlarged image signal SBn-1 'of a square pixel array having the same number of pixels as signal Bn-1'
Is obtained (step S24). Next, the processed band-limited image signal Bn-1 'and the enlarged image signal SBn-1' are added to obtain an addition signal San-1 (step S25). Then, it is determined whether an addition signal of the highest frequency band has been obtained (step S26), and step S2 is performed.
If affirmative, the process ends.

On the other hand, if step S26 is denied, the process returns to step S21, and from step S21 to step S2
Step 6 is repeated. Thereby, the addition signal Sa0 of the highest frequency band, that is, the processed image signal Sproc is obtained.

Returning to FIG. 16, the obtained processed image signal Sproc is reproduced as a visible image by the image output means 4 (step S4), and the process is terminated.

FIG. 19 is a diagram for explaining the difference in frequency characteristics between the band-limited image signal obtained by the conventional Laplacian pyramid method and the band-limited image signal obtained by the method of the present embodiment. 7A is a diagram illustrating frequency characteristics of a band-limited image signal obtained by the method of the Laplacian pyramid, and FIG. 7B is a diagram illustrating frequency characteristics of a band-limited image signal obtained by the method of the present embodiment. In FIG. 19, fs / 2 is the Nyquist frequency of the image signal of the square pixel array, and the solid line represents the division of the frequency band in which each band-limited image signal can be reproduced, that is, the maximum boundary of the band-limited image signal.

In the Laplacian pyramid method, since the band-limited image signal represents a detail image having a size of 1 / ^22N of the original image represented by the original image signal S0, the original image signal S0 is rectangular. The frequency characteristic of each band-limited image signal is rectangular as shown in FIG.

On the other hand, the band-limited image signal obtained according to the present embodiment includes a detail image having a size of 1/2 ^2N of the original image and an intermediate signal between 1/2 ^2N size and 1/2 ^{2 (N + 1)} size. Represents a detailed image of size. For this reason, the frequency band of the original image signal S0 is alternately divided into a rectangular shape and a rhombic shape, and as shown in FIG. 19B, the frequency characteristics of the rectangular and rhombic shapes also alternate in each band-limited image signal. Will appear.

Here, among the band-limited image signals Bk obtained according to the present embodiment, the maximum boundary of the frequency band of the band-limited image signal Bk in which k is an odd number and has a rectangular frequency characteristic is obtained by the Laplacian pyramid method. It is the same as the maximum boundary of the frequency band of each band-limited image signal obtained. On the other hand, the maximum boundary of the frequency band of the band-limited image signal Bk in which k is an even number and has a rhombic frequency characteristic is located between the maximum boundaries of the band-limited image signals obtained by the Laplacian pyramid method. . For this reason,
According to the present embodiment, it is possible to obtain a band-limited image signal Bk obtained by dividing the original image signal S0 into smaller frequency bands as compared with the conventional Laplacian pyramid method. Specifically, the band-limited image signal obtained by the Laplacian pyramid method represents a frequency band of 1/2 ^N to 1/2 ^{N + 1} of the Nyquist frequency fs / 2 of the original image signal S0. On the other hand, in the present embodiment,
Within one frequency band of the band-limited image signal obtained by the Laplacian pyramid technique, 1/2 ^N to 1/2 ^M and 1 / ^N of the Nyquist frequency fs / 2 of the original image signal S0.
This includes band-limited image signals of two frequency bands of 2 ^M to ⁺ 1/2 ^{N + 1} (N <M <N + 1).

Therefore, by performing the above-described enhancement processing on the band-limited image signal representing the detailed image for each of the plurality of frequency bands obtained by the present embodiment, image processing is performed at finer frequency band intervals. Therefore, the image quality can be controlled in more detail. As a result, the processed image signal Sproc obtained by subjecting the processed band-limited image signal Bk ′ to inverse multi-resolution conversion can reproduce a higher-quality processed image.

In the Laplacian pyramid method, the original image signal S0 is band-divided in a rectangular shape to obtain a band-limited image signal. In particular, it is desired to suppress a steep edge by using a nonlinear function of enhancement processing. In this case, edges can be favorably suppressed in the vertical and horizontal directions, that is, in the direction in which pixels are arranged in a square shape. However, since the band-limited image signal is not band-divided into a rhombus shape, edges cannot be satisfactorily suppressed in oblique directions, and as a result, edges are excessively emphasized in oblique directions. On the other hand, the band-limited image signal obtained by the present embodiment is obtained by dividing the band width of the original image into a square shape and a checkered shape alternately. If used, edges can be favorably suppressed even in oblique directions. Therefore, it is possible to prevent the edge from being excessively emphasized regardless of the direction of the edge.

Next, a second embodiment of the present invention will be described. In the first embodiment, the original image signal S0 having the square pixel array is subjected to the multi-resolution conversion and the enhancement process is performed, and then the inverse multi-resolution conversion is performed. The difference from the first embodiment is that the original image signal Ss0 is subjected to multi-resolution conversion. FIG. 20 is a diagram for explaining processing performed in the second embodiment of the present invention. In the second embodiment, since the original image signal Ss0 has a checkered pixel array, the original image signal Ss0 is first subjected to interpolation c, which is a checkerboard square reduction process, and compared with the original image signal Ss0. As a result, a reduced image signal Ss1 having a square pixel array having approximately half the number of pixels is obtained. The means for performing the interpolation processing c corresponds to the first means in the second multi-resolution conversion device according to the present invention.

Then, the reduced image signal Ss1 is subjected to an interpolation process d, which is a square checkerboard enlargement process, to obtain the original image signal Ss1.
An enlarged image signal SBs1 representing an enlarged image of a checkered pixel array having the same number of pixels as s0 is obtained. Then, by subtracting the enlarged image signal SBs1 from the original image signal Ss0, a band-limited image signal Bs1 of the highest frequency band is obtained.
The means for performing the interpolation processing d corresponds to the second means in the second multi-resolution conversion device according to the present invention.

On the other hand, the reduced image signal Ss1 obtained by the interpolation process c is subjected to an interpolation process a, which is a square checkerboard reduction process, and is compared with the reduced image signal Ss1 by about 1 /
Reduced image signal Ss of a checkered pixel array having two pixels
Get 2. The means for performing the interpolation processing a corresponds to the fourth means in the second multi-resolution conversion device according to the present invention. The reduced image signal Ss2 has half the number of vertical and horizontal pixels, that is, one fourth the number of pixels, compared to the original image signal Ss0. Therefore, the size of the image represented by the reduced image signal Ss2 is 1/2 ² of the original image.

Next, the reduced image signal Ss2 is subjected to an interpolation process b as a checkerboard square enlargement process to obtain an enlarged image signal SBs2 having a square pixel array having the same number of pixels as the reduced image signal Ss1. The means for performing the interpolation processing b corresponds to the fifth means in the second multi-resolution conversion device according to the present invention. Then, by subtracting the enlarged image signal SBs2 from the reduced image signal Ss1, a band-limited image signal Bs2 having a frequency band one step lower than the highest frequency band is obtained.

Then, interpolation processing c, d, a, and b are performed on the reduced image signal Ss2 in the same manner as described above, and a reduced image signal Ssk (k = 1 to n) obtained by the interpolation processing is obtained.
By obtaining a difference value from the enlarged image signal SBsk (k = 1 to n), a band-limited image signal Bsk (k = 1 to n) for each of a plurality of frequency bands is obtained. Here, when k is an odd number, the reduced image represented by the reduced image signal Ssk has a square pixel array, and the enlarged image represented by the enlarged image signal SBk and the detailed image represented by the band-limited image signal Bk are checkered. Pixel arrangement. On the other hand, when k is an even number, the reduced image represented by the reduced image signal Ssk has a checkered pixel array, and the enlarged image represented by the enlarged image signal SBsk and the detailed image represented by the band-limited image signal Bsk are square. Pixel arrangement. When the band-limited image signal Bsn of the lowest frequency band is obtained by using the reduced image signal Ssn obtained by performing the interpolation processing a, the detailed image represented by the band-limited image signal Bsn is a square pixel. It becomes an array. On the other hand, when the band-limited image signal Bsn of the lowest frequency band is obtained using the reduced image signal Ssn obtained by performing the interpolation processing c, the detailed image represented by the band-limited image signal Bsn is a checkered pixel. It becomes an array.

The emphasis processing means 2 'performs emphasis processing on the band-limited image signal Bsk in the same manner as the emphasis processing means 2 in the first embodiment to obtain a processed band-limited image signal Bsk'.

The inverse multi-resolution conversion means 3 'performs inverse multi-resolution conversion on the processed band-limited image signal Bsk' as described below, and outputs the processed image signal Sspr.
get oc. Here, it is assumed that the processing up to the interpolation processing a is performed to obtain the reduced image signal Ssn of the lowest frequency band representing the reduced image in which the pixels are arranged in a checkered pattern. In this case, the detailed image represented by the band-limited image signal Bsn of the lowest frequency band has pixels arranged in a square shape.
First, for the reduced image signal Ssn of the lowest frequency band,
An interpolation process b, which is a checkerboard square enlargement process, is performed, and an enlarged image signal SBsn having a square pixel array is obtained. This enlarged image signal SBsn is the same as the enlarged image signal SBsn obtained by the multi-resolution conversion means 1 '. Then, by adding the processed band-limited image signal Bsn 'of the lowest frequency band of the square pixel array and the enlarged image signal SBsn, the addition signal Sas of the square pixel array is obtained.
n-1 is obtained.

Next, interpolation processing d, which is square checkerboard enlargement processing, is performed on the addition signal Sasn-1 to obtain an enlarged image signal SBsn-1 'having a checkerboard pixel array. The enlarged image signal SBsn-1 'is a processed band-limited image signal Bsn-1' in a frequency band one step higher than the lowest frequency band.
And an addition signal Sasn-2 is obtained. Hereinafter, the interpolation processing d (k: odd number,
In the case of an even number, the addition of the enlarged image signal SBsk 'obtained by the interpolation processing b) and the interpolation processing d and b and the processed band-limited image signal Bsk' is performed on the enlarged image signal SBs1 'of the highest frequency band and the addition of the highest frequency band. This processing is repeated until the addition with the processed band-limited image signal Bs1 'is performed, thereby obtaining an addition signal Sas0 of the highest frequency band. The sum signal Sas0 of the highest frequency band is a processed image signal Ssproc.
Becomes

As described above, according to the second embodiment, it is possible to obtain a band-limited image signal in which the band is divided more finely as compared with the conventional Laplacian pyramid method, thereby controlling the image quality in more detail. can do. Therefore, the processed band-limited image signal Bs
The processed image signal Ssproc obtained by performing the inverse multi-resolution conversion of k ′ can reproduce a processed image of higher quality.

In the first and second embodiments, the band-limited image signals Bk and Bsk are enhanced by using the table T representing the nonlinear function shown in FIG. A non-linear function as shown in FIG. As shown in FIG. 21, this table T suppresses the signal when the pixel value is larger than the threshold value α, and does not enhance or suppress the signal when the pixel value is smaller than the threshold value β. It is. By using such a table T, in the band limited image signals Bk and Bsk, a portion having a very small pixel value which can be regarded as noise is not emphasized. Therefore,
By performing inverse multi-resolution conversion of the processed band-limited image signals Bk 'and Bsk' processed using such a table T1, a processed image signal Sproc capable of reproducing a high-quality image with no noticeable noise. , Ssproc.

[Brief description of the drawings]

FIG. 1 is a schematic block diagram illustrating a configuration of an image processing system including a multi-resolution conversion device and an image processing device according to an embodiment of the present invention.

FIG. 2 is a diagram for explaining processing performed in the first embodiment;

FIG. 3 is a diagram showing a pixel array of an original image signal.

FIG. 4 is a diagram showing a filter used in interpolation processing a;

FIG. 5 is a diagram for explaining a filtering process in an interpolation process a;

FIG. 6 is a diagram showing a checkered pixel array of a reduced image signal.

FIG. 7 is a diagram illustrating a filter used in an interpolation process b;

FIG. 8 is a diagram showing a square pixel array of an enlarged image signal;

FIG. 9 is a diagram showing a filter used in the interpolation processing c.

FIG. 10 is a diagram illustrating a filtering process in an interpolation process c.

FIG. 11 is a diagram showing a square pixel array of a reduced image signal;

FIG. 12 is a diagram showing a filter used in the interpolation processing d.

FIG. 13 is a diagram showing a checkered pixel array of an enlarged image signal;

FIG. 14 is a diagram showing an example of a non-linear function used as a table stored in a table storage unit.

FIG. 15 is a diagram schematically showing multi-resolution conversion, enhancement processing, and inverse multi-resolution conversion performed in the first embodiment.

FIG. 16 is a flowchart showing the operation of the first embodiment;

FIG. 17 is a flowchart showing the operation of a multi-resolution conversion process;

FIG. 18 is a flowchart showing the operation of the inverse multi-resolution conversion process;

FIG. 19 is a view for explaining a difference in frequency characteristics between a band-limited image signal obtained by the conventional Laplacian pyramid method and a band-limited image signal obtained by the method of the present embodiment.

FIG. 20 is a view for explaining processing performed in the second embodiment of the present invention;

FIG. 21 is a diagram showing another example of a nonlinear function used as a table.

FIG. 22 is a diagram showing a square pixel array;

FIG. 23 is a diagram showing a checkered pixel array;

[Explanation of symbols]

 1, 1 'multi-resolution conversion means 2, 2' enhancement processing means 3, 3 'inverse multi-resolution conversion means 4, 4' image output means

──────────────────────────────────────────────────続 き Continued on the front page (51) Int.Cl. ⁷ Identification symbol FI Theme coat ゛ (Reference) G09G 5/391 H04N 7/18 U 5L096 5/00 550 G09G 5/00 520V 5/36 5/36 520J H04N 1/409 520F 7/18 520G H04N 1/40 101D 101C F-term (reference) 5B057 CA08 CA12 CA16 CB08 CB12 CB16 CC02 CD05 CE02 CE06 CG05 5C054 EA05 ED14 EG01 EJ04 FC12 HA12 5C076 AA21 AA22 PP04 CB02 PP47 PP49 PQ23 5C082 AA04 BA12 BA35 BB46 CA22 CA33 CA34 CA85 CB03 DA87 MM02 MM05 MM10 5L096 AA06 EA03 EA33 FA22 GA05 GA06

Claims (18)

[Claims] 1. A multi-resolution conversion process is performed on an original image signal having a square pixel array to obtain a band-limited image signal representing an image of each of a plurality of frequency bands of an original image represented by the original image signal. A first step of obtaining a first reduced image signal by applying a square checkerboard reduction process to the original image signal; a checkerboard square enlargement process for the first reduced image signal; And a second step of obtaining a first band-limited image signal by obtaining a difference value between the original image signal and the first enlarged image signal.
A fourth step of performing a checkerboard square reduction process on the first reduced image signal to obtain a second reduced image signal, and performing a square checkerboard enlargement process on the second reduced image signal A fifth step of obtaining a second enlarged image signal by calculating a difference value between the first reduced image signal and the second enlarged image signal, so that the difference value is one more than the first band-limited image signal. A sixth step of obtaining a second band-limited image signal of a stepped low frequency band, and the first to third steps for the second reduced image signal until a band-limited image signal of a desired frequency band is obtained. Or a seventh step of repeating the first to sixth steps. 2. A multi-resolution conversion process is performed on an original image signal having a checkered pixel array to obtain a band-limited image signal representing an image of each of a plurality of frequency bands of the original image represented by the original image signal. A first step of obtaining a first reduced image signal by performing a checkerboard square reduction process on the original image signal, a square checkerboard enlargement process on the first reduced image signal. And a second step of obtaining a first band-limited image signal by obtaining a difference value between the original image signal and the first enlarged image signal.
A fourth step of obtaining a second reduced image signal by performing a square checkerboard reduction process on the first reduced image signal, and performing a checkerboard square enlargement process on the second reduced image signal A fifth step of obtaining a second enlarged image signal by calculating a difference value between the first reduced image signal and the second enlarged image signal, so that the difference value is one more than the first band-limited image signal. A sixth step of obtaining a second band-limited image signal of a stepped low frequency band, and the first to third steps for the second reduced image signal until a band-limited image signal of a desired frequency band is obtained. Or a seventh step of repeating the first to sixth steps. 3. A band-limited image signal of a predetermined frequency band among band-limited image signals representing images of the plurality of frequency bands obtained by the multi-resolution conversion method according to claim 1 or 2. An image processing method comprising performing predetermined image processing. 4. The image processing method according to claim 3, wherein the predetermined image processing is an enhancement processing. 5. The image processing method according to claim 3, wherein the predetermined image processing is noise suppression processing. 6. A processed image signal is obtained by performing an inverse multi-resolution conversion process on the band-limited image signal subjected to the predetermined image processing and the other band-limited image signal. The image processing method according to claim 3. 7. A multi-resolution conversion process is performed on an original image signal having a square pixel array to obtain a band-limited image signal representing an image of each of a plurality of frequency bands of the original image represented by the original image signal. A first means for obtaining a first reduced image signal by subjecting the original image signal to a square checkerboard reduction process, a checkerboard square enlargement process for the first reduced image signal. A second means for obtaining a first enlarged image signal by performing the following; a third means for obtaining a first band-limited image signal by obtaining a difference value between the original image signal and the first enlarged image signal
Means for obtaining a second reduced image signal by performing a checkerboard square reduction process on the first reduced image signal, and performing a square checkerboard enlargement process on the second reduced image signal Fifth means for obtaining a second enlarged image signal, thereby obtaining a difference value between the first reduced image signal and the second enlarged image signal, thereby obtaining one more than the first band-limited image signal. A sixth means for obtaining a second band-limited image signal of a stepped low frequency band, wherein the first to the second to the second reduced image signal are obtained until a band-limited image signal of a desired frequency band is obtained. 3. A multi-resolution conversion device, comprising: multi-resolution conversion means for repeating the processing in the third means or the processing in the first to sixth means. 8. A multi-resolution conversion process is performed on an original image signal having a checkered pixel array to obtain a band-limited image signal representing an image of each of a plurality of frequency bands of the original image represented by the original image signal. A first means for obtaining a first reduced image signal by performing a checkerboard square reduction process on the original image signal, comprising: a square checkerboard enlargement process for the first reduced image signal. A second means for obtaining a first enlarged image signal by performing the following; a third means for obtaining a first band-limited image signal by obtaining a difference value between the original image signal and the first enlarged image signal
Means for obtaining a second reduced image signal by performing a square checkerboard reduction process on the first reduced image signal, and performing a checkerboard square enlargement process on the second reduced image signal Fifth means for obtaining a second enlarged image signal, thereby obtaining a difference value between the first reduced image signal and the second enlarged image signal, thereby obtaining one more than the first band-limited image signal. A sixth means for obtaining a second band-limited image signal of a stepped low frequency band, wherein the first to the second to the second reduced image signal are obtained until a band-limited image signal of a desired frequency band is obtained. 3. A multi-resolution conversion device, comprising: multi-resolution conversion means for repeating the processing in the third means or the processing in the first to sixth means. 9. A band-limited image signal of a predetermined frequency band among band-limited image signals representing images for each of the plurality of frequency bands obtained by the multi-resolution conversion device according to claim 7. An image processing apparatus, comprising: an image processing unit that performs predetermined image processing. 10. The image processing apparatus according to claim 9, wherein the predetermined image processing is an enhancement processing. 11. The image processing apparatus according to claim 9, wherein the predetermined image processing is a noise suppression processing. 12. An inverse multi-resolution conversion unit for performing an inverse multi-resolution conversion process on the band-limited image signal subjected to the predetermined image processing and the other band-limited image signal to obtain a processed image signal. The image processing apparatus according to claim 9, further comprising: 13. A multi-resolution conversion process is performed on an original image signal having a square pixel array to obtain a band-limited image signal representing an image of each of a plurality of frequency bands of the original image represented by the original image signal. A computer-readable recording medium recording a program for causing a computer to execute a multi-resolution conversion method, wherein the program performs a square checkerboard reduction process on the original image signal to generate a first reduced image signal. The first to get
A second procedure of obtaining a first enlarged image signal by applying a checkerboard square enlargement process to the first reduced image signal; a difference value between the original image signal and the first enlarged image signal To obtain a first band-limited image signal.
A fourth procedure of performing a checkerboard square reduction process on the first reduced image signal to obtain a second reduced image signal; performing a square checkerboard enlargement process on the second reduced image signal A fifth procedure for obtaining a second enlarged image signal by calculating the difference value between the first reduced image signal and the second enlarged image signal, and A sixth procedure for obtaining a second band-limited image signal of a stepped low frequency band, and the first to third steps for the second reduced image signal until a band-limited image signal of a desired frequency band is obtained. Or a seventh procedure for repeating the first to sixth procedures. 14. A multi-resolution conversion process is performed on an original image signal having a checkered pixel array to obtain a band-limited image signal representing an image of each of a plurality of frequency bands of the original image represented by the original image signal. A computer-readable recording medium recording a program for causing a computer to execute a multi-resolution conversion method, wherein the program performs a first reduction image signal by performing a checkerboard square reduction process on the original image signal. The first to get
A second procedure of obtaining a first enlarged image signal by performing a square checkerboard enlargement process on the first reduced image signal; a difference value between the original image signal and the first enlarged image signal To obtain a first band-limited image signal.
A fourth step of performing a square checkerboard reduction process on the first reduced image signal to obtain a second reduced image signal, and performing a checkerboard square enlargement process on the second reduced image signal A fifth procedure for obtaining a second enlarged image signal by calculating the difference value between the first reduced image signal and the second enlarged image signal, and A sixth procedure for obtaining a second band-limited image signal of a stepped low frequency band, and the first to third steps for the second reduced image signal until a band-limited image signal of a desired frequency band is obtained. Or a seventh procedure for repeating the first to sixth procedures. 15. A band-limited image signal of a predetermined frequency band among band-limited image signals representing images of the plurality of frequency bands obtained by the multi-resolution conversion method according to claim 1 or 2. A computer-readable recording medium in which a program for causing a computer to execute an image processing method having a procedure for performing predetermined image processing is recorded. 16. The computer-readable recording medium according to claim 15, wherein said predetermined image processing is enhancement processing. 17. The computer-readable recording medium according to claim 15, wherein the predetermined image processing is a noise suppression processing. 18. A process for obtaining a processed image signal by performing an inverse multi-resolution conversion process on the band-limited image signal subjected to the predetermined image processing and the other band-limited image signal. The computer-readable recording medium according to any one of claims 15 to 17, wherein: