JP6955386B2 - Super-resolution device and program - Google Patents

Description

The present invention relates to a super-resolution device and a program that super-resolution magnifies the resolution of an input image by two octaves or more.

Conventionally, there is known a spatial super-resolution technique in which an image is subjected to spatial super-resolution processing in the spatial direction to generate an image having a higher resolution than the original image. Learning type and reconstruction type are known as spatial super-resolution technology (see, for example, Non-Patent Document 1).

The learning type is a method in which a set of a low frequency component and a high frequency component is held as a database, and the high frequency component is set as a super-resolution component by matching the original image and the low frequency component. However, there is a problem that a huge database is required to super-resolution natural images and the like with high image quality. Further, since iterative calculation is required, real-time processing becomes difficult.

On the other hand, the reconstruction type is a method of generating a super-resolution component by linear / non-linear filtering or registration between a plurality of frames. Here, it has been difficult to generate super-resolution high-frequency components with high accuracy and accuracy by linear / nonlinear filtering. Therefore, for example, in registration super-resolution by aligning a plurality of low-resolution images, a method has been proposed that enables robust super-resolution processing even when an object is obscured (for example, non-resolution). See Patent Document 2).

Masatoshi Okutomi, Masayuki Tanaka, Hidenori Takeshima, Nobuyuki Matsumoto, "Latest Trends in Image Super-Resolution Processing Technology", Journal of the Institute of Electronics, Information and Communication Engineers, Vol. 93, No. 8, pp. 693-698, 2010. Masayuki Tanaka, Yoichi Yaguchi, Eiji Yoshikawa, Masatoshi Okutomi, "Robust super-resolution processing based on pixel selection considering the amount of misalignment", IEICE Transactions D, Vol. J92-D, No. 5, pp. 650-660, May 2009.

However, in the conventional multi-frame registration process, when the registration accuracy and accuracy are high, a super-resolution high-frequency component with high accuracy and accuracy can be generated, but when the super-resolution is enlarged by two octaves or more, Since it is necessary to perform registration even from an image frame at a considerably distant time position, there is a problem that it is difficult to generate a super-resolution high-frequency component with high accuracy and accuracy.

An object of the present invention made in view of such circumstances is to provide a super-resolution device and a program capable of performing super-resolution enlargement of 2 octaves or more with high accuracy and high accuracy.

In order to solve the above problems, the super-resolution device according to the present invention is a super-resolution device that generates a super-resolution image obtained by enlarging the original image by n octaves, and has an n-th order frequency with respect to the original image. A frequency decomposition unit that performs decomposition processing to generate a frequency decomposition image, a reference source block in the original image, and a reference destination similar to the reference source block in the n-th order low frequency component image generated by the frequency decomposition unit. An alignment unit that generates alignment information indicating the positional relationship with the block and an integer from 0 to n-1 are k, and the (nk) order generated by the frequency decomposition unit according to the alignment information. The octave allocation part that allocates the high-frequency component image to the (k + 1) octave super-resolution high-frequency component to generate the (k + 1) octave super-resolution high-frequency component image, and the size of the reference source block are doubled in the horizontal and vertical directions. The similarity between the first block and the second block having the same phase position as the referenced block in the (n-1) order low frequency image LL ^{n-1 is calculated, and only when the similarity exceeds the threshold value.} , The similarity determination unit that outputs the alignment information to the octave allocation unit, the n-th order frequency reconstruction process using the original image as a low frequency component and the (k + 1) octave super-resolution high-frequency component image as a high-frequency component. It is characterized by including a frequency reconstruction unit that generates a super-resolution image.

In order to solve the above problems, the super-resolution device according to the present invention is a super-resolution device that generates a super-resolution image obtained by enlarging the original image by n octaves, and has an n-th order frequency with respect to the original image. A frequency decomposition unit that performs decomposition processing to generate a frequency decomposition image, a reference source block in the original image, and a reference destination similar to the reference source block in the n-th order low frequency component image generated by the frequency decomposition unit. An alignment unit that generates alignment information indicating the positional relationship with the block and an integer from 0 to n-1 are k, and the (nk) order generated by the frequency decomposition unit according to the alignment information. The octave allocation part that allocates the high-frequency component image to the (k + 1) octave super-resolution high-frequency component to generate the (k + 1) octave super-resolution high-frequency component image, and the size of the reference source block are doubled in the horizontal and vertical directions. A second block group consisting of the first block, a second block having the same phase position as the referenced block in the (n-1) order low frequency image LL ^{n-1, and a block having a phase shift of the second block.} The phase adjustment unit that calculates the similarity with the second block group and includes the phase information indicating the position of the block having the maximum similarity with the first block in the alignment information, and the original image at a low frequency. It is characterized by including, as a component, a frequency reconstruction unit that generates a super-resolution image by performing n-th order frequency reconstruction processing using the (k + 1) octave super-resolution high-frequency component image as a high-frequency component .

Further, in the super-resolution device according to the present invention, the phase adjusting unit outputs the alignment information to the octave allocation unit only when the similarity of the block having the maximum similarity exceeds the threshold value. It is a feature.

Further, in order to solve the above-mentioned problems, the program according to the present invention is characterized in that the computer functions as the above-mentioned super-resolution device.

According to the present invention, it is possible to perform super-resolution enlargement of 2 octaves or more with high accuracy and high accuracy.

It is a block diagram which shows the structural example of the super-resolution apparatus which concerns on 1st Embodiment of this invention. It is a figure which shows the outline of the processing of the alignment part in the super-resolution apparatus which concerns on 1st Embodiment of this invention. It is a figure which shows the outline of the processing of the octave allocation part in the super-resolution apparatus which concerns on 1st Embodiment of this invention. It is a block diagram which shows the structural example of the super-resolution apparatus which concerns on 2nd Embodiment of this invention. It is a figure which shows the outline of the process of the similarity determination part of the super-resolution apparatus which concerns on 2nd Embodiment of this invention. It is a block diagram which shows the structural example of the super-resolution apparatus which concerns on 3rd Embodiment of this invention. It is a figure which shows the outline of the processing of the phase adjustment part of the super-resolution apparatus which concerns on 3rd Embodiment of this invention.

Hereinafter, embodiments of the present invention will be described in detail with reference to the drawings.

(First Embodiment)
FIG. 1 shows a configuration example of the super-resolution device according to the first embodiment of the present invention. The super-resolution device 1 shown in FIG. 1 includes a frequency decomposition unit 11, an alignment unit 12, an octave allocation unit 13, and a frequency reconstruction unit 14.

The super-resolution device 1 inputs an original image and performs super-resolution processing to generate a super-resolution image that is super-resolution-enlarged by two octaves or more. When the super-resolution image is enlarged by n octaves, the size of the super-resolution image becomes ^{2 n times each in the vertical and horizontal directions of the original image.}

The frequency decomposition unit 11 performs frequency decomposition (multi-resolution decomposition) processing of a plurality of layers on the input original image to generate a frequency decomposition image. When the n-th order frequency decomposition is performed on the original image, the frequency decomposition unit 11 outputs the n-th order low frequency component image LL ⁿ to the alignment unit 12 and outputs the high frequency component image of the original image to the octave allocation unit 13. .. When the decomposition rank n is 2, the high-frequency component image is composed of the first-order high-frequency component images LH ¹ , HL ¹ , HH ^1, and the second-order high-frequency component images LH ² , HL ² , and HH ² .

The frequency decomposition unit 11 performs wavelet decomposition processing using a wavelet (for example, CDF9 / 7) having linear phase and having a tap length equal to or larger than the threshold value in consideration of the discontinuity of the gradation value of the original image. You may do it.

The alignment unit 12 is ^{a position indicating the positional relationship between the reference source block in the original image LL 0} and the reference destination block similar to the reference source block in the ^n-th order low frequency component image LL n generated by the frequency decomposition unit 11. The alignment information is generated and output to the octave allocation unit 13.

FIG. 2 is a diagram showing an outline of the alignment process in the alignment unit 12. The alignment unit 12 ^{performs block matching between both frames using the original image LL 0} as a reference frame and the nth-order low-frequency component image LL ⁿ (n = 2 in FIG. 2) as a reference frame, and is similar within the search range. Generates alignment information that indicates the positional relationship of the blocks with the highest degree (correlation). Block matching is performed by a known method using evaluation functions such as sum of absolute error (SAD) and sum of squared error (SSD). Further, the block matching is performed with a decimal pixel accuracy by, for example, interpolation processing using the parabolic fitting function shown in the equation (1). If the evaluation function value of SAD or SSD exceeds the threshold value, it may not be adopted as the alignment information.

Here, when the pixel position at the search position is x, the SSD (x) represents the SSD value at the pixel position, and more specifically, the SSD (0) is the center position (the position where the SSD value is minimized). SSD (-1) is the SSD value at the position of the adjacent pixel in the −x direction or −y direction from the center position, and SSD (1) is the SSD value at the position of the adjacent pixel in the + x direction or + y direction from the center position. Represents. From the equation (1), the pixel positions (decimal pixel positions) with the decimal pixel accuracy in the horizontal or vertical direction can be calculated respectively.

The octave allocation unit 13 first sets an initial value in order to estimate a high-frequency component (super-resolution high-frequency component image) of the super-resolution image that exceeds the Nyquist frequency of the original image. The super-resolution high-frequency component image is a super-resolution horizontal high-frequency component image which is a horizontal high-frequency component of the super-resolution image, a super-resolution vertical high-frequency component image which is a vertical high-frequency component of the super-resolution image, and a super-resolution image. It consists of a super-resolution diagonal high-frequency component image, which is a diagonal high-frequency component of. For example, as an initial value, the values of all pixels of the super-resolution horizontal high-frequency component image, the super-resolution vertical high-frequency component image, and the super-resolution diagonal high-frequency component image are set to 0.

Next, assuming that an integer from 0 to n-1 is k, the octave allocation unit 13 obtains a (nk) order high frequency component image exceeding (k + 1) octave according to the alignment information generated by the alignment unit 12. An octave super-resolution high-frequency component image is generated by allocating to the resolution high-frequency component (k + 1), and is output to the frequency reconstruction unit 14. Here, when allocating the (nk) order high frequency component image, it is assumed that the alignment information of the same phase position in the ^{nth order low frequency component image LL n is followed.} This is because if ^{a block P in the original image LL 0 is similar to the block Q in the nth} order low frequency component image LL n, it has the same phase as the block P in the unknown (k + 1) octave super-resolution image. This is because the position block is likely to be similar to the block at the same phase position as the block Q in the (nk) order high frequency component image.

With reference to FIG. 3, the processing of the octave allocation unit 13 when the decomposition rank n is 2 will be described. When n = 2, k = 0,1. The octave allocation unit 13 sets a ^{block in the second-order horizontal high-frequency component image LH 2 at the} same phase position as the reference block based on the alignment information between the ^{original image LL 0} and the second-order low-frequency component image LL ^{2. , The 1-octave super-resolution horizontal high-frequency component image LH 0} is generated by allocating to the 1-octave super-resolution horizontal high-frequency component at the same phase position as the reference source block. Similarly, a 1-octave super-resolution vertical high-frequency component image HL ⁰ and a 1-octave super-resolution diagonal component image HH ⁰ are generated. Further, the octave allocation unit 13 is in ^{the first-order horizontal high-frequency component image LH 1 at the} same phase position as the reference block based on the alignment information between the ^{original image LL 0} and the second-order low-frequency component image LL ² . By allocating the block to the 2-octave super-resolution horizontal high-frequency component at the same phase position as the reference block, the 2-octave super-resolution horizontal high-frequency component image LH ^-1 is generated. Similarly, a 2-octave super-resolution vertical high-frequency component image HL ^-1 and a 2-octave super-resolution diagonal component image HH ^-1 are generated. Comparing the sizes of the blocks in the first-order high-frequency component image and the blocks in the second-order low-frequency component image LL ^{2 at} the same phase position, the size of the blocks in the first-order high-frequency component image is the second-order low-frequency component image. It is doubled in the horizontal and vertical directions with respect to the block of ^{LL 2.}

Further, when the positional relationship of similar blocks is obtained with the fractional pixel accuracy in the alignment unit 12, the octave allocation unit 13 super-resolutions after allocation in order to align the fractional pixel position with the normal pixel position. Interpolation using a point spread function (PSF) that simulates the resolution deterioration process of the optical system for horizontal high-frequency component images, super-resolution vertical high-frequency component images, and super-resolution diagonal high-frequency component images. conduct. Equation (2) shows the point spread function. Here, w is the half width (variance value) of the Gaussian function.

When there are a plurality of candidates as horizontal, vertical, and diagonal super-resolution high-frequency components, the octave allocation unit 13 averages those values or reconstructs the maximum a posteriori (MAP). Estimate an unknown value. For more information on MAP reconstruction, see, for example, E. Levitan and G. Herman: “A maximum a posteriori probability expectation maximization algorithm for image reconstruction in emission tomography”, IEEE Transactions on Medical Imaging, vol. 6, no. 3, pp. . 185-192, Sep. 1987. Further, as another method, the super-resolution high-frequency component image may be estimated by the ML method or the weighting according to the distance of the allocated pixels.

The frequency reconstruction unit 14 uses the original image as a low frequency component and the (k + 1) octave super-resolution high-frequency component image allocated by the octave allocation unit 13 as a high-frequency component, and performs n-th order frequency reconstruction processing to obtain a super-resolution image. Is generated and output to the outside. When the frequency decomposition unit 11 performs the wavelet decomposition processing as the frequency decomposition processing, the frequency reconstruction unit 14 performs the wavelet reconstruction processing using the same wavelet. In the example of FIG. 3, by performing the second-order wavelet reconstruction processing, a 2-octave super-resolution image in which the number of pixels in the horizontal direction is doubled and the number of pixels in the vertical direction is doubled with respect to the original image is generated. ..

A computer can be preferably used to function as the above-mentioned super-resolution device 1, and such a computer provides a program describing processing contents for realizing each function of the super-resolution device 1. It can be realized by storing it in a storage unit and reading and executing this program by the CPU of the computer. This program can be recorded on a computer-readable recording medium.

As described above, the invention according to the first embodiment provides alignment information indicating the positional relationship between the reference source block in the original image and the reference destination block similar to the reference source block in the n-th order low frequency component image. Then, according to the alignment information, the (nk + 1) order high-frequency component image is assigned to the (k + 1) octave super-resolution high-frequency component to generate the (k + 1) octave super-resolution high-frequency component image. This makes it possible to perform super-resolution enlargement of n octaves with high accuracy and high accuracy.

(Second Embodiment)
Next, a second embodiment of the present invention will be described. FIG. 4 shows a configuration example of the super-resolution device according to the second embodiment of the present invention. The super-resolution device 2 shown in FIG. 4 includes a frequency decomposition unit 11, an alignment unit 12, an octave allocation unit 13, a frequency reconstruction unit 14, and a similarity determination unit 15. The super-resolution device 2 is different from the super-resolution device 1 according to the first embodiment in that it further includes a similarity determination unit 15.

Similar to the first embodiment, the frequency decomposition unit 11 performs frequency decomposition processing of a plurality of layers on the input original image to generate a frequency decomposition image. Then, the n-th-order low-frequency component image LL ⁿ is output to the alignment unit 12, the high-frequency component image is output to the octave allocation unit 13, and the (n-j) -th-order low-frequency component image LL ^n-j is output to the similarity determination unit. Output to 15. Here, j is an integer from 1 to n-1.

Similar to the first embodiment, the alignment unit 12 determines ^{the positional relationship between the reference source block in the original image LL 0} and the reference destination block similar to the reference source block in the n-th order low frequency component image LL ^n. The indicated alignment information is generated and output to the similarity determination unit 15.

The similarity determination unit 15 determines the alignment information in order to improve the accuracy of the octave allocation unit 13. Specifically, in the first ^{block in which the size of the reference source block in the original image LL 0} is doubled in the horizontal and vertical directions, and in the (n-1) order low frequency image LL ^n-1 , the nth order is low. The similarity is calculated for the second block at the same phase position as the reference block in the frequency image LL ⁿ , and the alignment information input from the alignment unit 12 is allocated in octaves only when the similarity exceeds the threshold value. Output to unit 13. If the similarity exceeds the threshold, the components in the high-frequency component images LH ^n-1 , HL ^n-1 , and HH ^{n-1 at} the same spatial phase position are unknown super-resolution horizontal exceeding the Nyquist frequency of the original image LL ^0. It can be assumed that the high-frequency component LH ^-1 , the super-resolution vertical high-frequency component HL ^-1 , and the super-resolution diagonal high-frequency component HH ^{-1 are similar.} When the similarity is calculated by SSD or SAD, the smaller the value, the higher the similarity.

With reference to FIG. 5, the processing of the similarity determination unit 15 when the decomposition rank n is 2 will be described. When n = 2, the similarity determination unit 15 determines that ^{the reference source block in the original image LL 0} has the size doubled in the horizontal and vertical directions in the first block A and the first-order low-frequency image LL ¹ . The degree of similarity with the second block B at the same phase position as the reference block in the second-order low-frequency image LL ^{2 is obtained.} Then, it is determined whether or not the similarity exceeds the threshold value. In the first-order low-frequency image LL ¹ , the size of the block at the same phase position as the reference block in the second-order low-frequency image LL ² is doubled in the horizontal and vertical directions with respect to the reference block. Therefore, the size of the reference source block in the original image LL ^{0 to be} compared is also doubled in the horizontal and vertical directions.

Further, when n is 3 or more, the similarity determination unit 15 sets the size of the reference source block in the ^{original image LL 0 in the horizontal and vertical directions, where j is an integer from 1 to n-1.} a first block was 2 ^j times, in (n-j) Kaihiku frequency image LL ^n-j, for the second block of the same phase position as the reference block in the n Kaihiku frequency image LL ⁿ similarity The alignment information input from the alignment unit 12 may be output to the octave allocation unit 13 only when the calculation is performed and the similarity exceeds the threshold value.

The octave allocation unit 13 allocates the (nk) order high frequency component image to the (k + 1) octave super-resolution high frequency component according to the alignment information input from the similarity determination unit 15 as in the first embodiment. (K + 1) Octave super-resolution high-frequency component image is generated and output to the frequency reconstruction unit 14.

Similar to the first embodiment, the frequency reconstruction unit 14 uses the original image as a low frequency component and the (k + 1) octave super-resolution high-frequency component image allocated by the octave allocation unit 13 as a high-frequency component, and uses the n-th order frequency. Reconstruction processing is performed to generate an n-octave super-resolution image, which is output to the outside.

A computer can be preferably used to function as the above-mentioned super-resolution device 2, and such a computer provides a program describing processing contents for realizing each function of the super-resolution device 2. It can be realized by storing it in a storage unit and reading and executing this program by the CPU of the computer. This program can be recorded on a computer-readable recording medium.

As described above, the invention according to the second embodiment describes the reference source block in the first block whose size is doubled in the horizontal and vertical directions, and in the (n-1) phase low frequency image LL ^n-1 . The similarity is calculated for the second block having the same phase position as the reference block, and the allocation is performed using the alignment information only when the similarity exceeds the threshold value. As a result, the allocation can be performed using only the highly accurate alignment information, so that it is possible to perform super-resolution enlargement of n octaves with high accuracy and high accuracy.

(Third Embodiment)
Next, a third embodiment of the present invention will be described. FIG. 6 shows a configuration example of the super-resolution device according to the third embodiment of the present invention. The super-resolution device 3 shown in FIG. 6 includes a frequency resolution unit 11, an alignment unit 12, an octave allocation unit 13, a frequency reconstruction unit 14, and a phase adjustment unit 16. The super-resolution device 2 is different from the super-resolution device 1 according to the first embodiment in that it further includes a phase adjusting unit 16.

Similar to the first embodiment, the frequency decomposition unit 11 performs frequency decomposition processing of a plurality of layers on the input original image to generate a frequency decomposition image. Then, the n-th-order low-frequency component image LL ⁿ is output to the alignment unit 12, the high-frequency component image is output to the octave allocation unit 13, and the (n-j) -th-order low-frequency component image LL ^n-j is output to the phase adjustment unit 16. Output to.

Similar to the first embodiment, the alignment unit 12 determines ^{the positional relationship between the reference source block in the original image LL 0} and the reference destination block similar to the reference source block in the n-th order low frequency component image LL ^n. The indicated alignment information is generated and output to the phase adjustment unit 16.

The phase adjusting unit 16 adjusts the phase in order to improve the accuracy of the octave allocation unit 13. Specifically, in the first ^{block in which the size of the reference source block in the original image LL 0} is doubled in the horizontal and vertical directions, and in the (n-1) order low frequency image LL ^n-1 , the nth order is low. The similarity was calculated for the second block having the same phase position as the reference block in the frequency image LL ⁿ and the second block group in which the second block was shifted by one pixel each in the horizontal, vertical, and diagonal directions. The phase information indicating the position of the block having the maximum similarity with the first block in the second block group is included in the alignment information and output to the octave allocation unit 13.

The processing of the phase adjusting unit 16 when the decomposition rank n is 2 will be described with reference to FIG. 7. When n = 2, the phase adjusting unit 16 performs ^{2 in the first block A in which the size of the reference source block in the original image LL 0} is doubled in the horizontal and vertical directions, and in the first-order low-frequency image LL ¹ . The second block B1 and the second block B1 at the same phase position as the reference block in the low-frequency image LL ² are shifted by one pixel in the horizontal direction, and the second block B2 and the second block B1 are shifted by one pixel in the vertical direction. The degree of similarity is obtained for each of the second block B3 and the second block B4 in which the second block B1 is shifted by one pixel in the diagonal direction (that is, one pixel each in the horizontal and vertical directions). Then, among the second block groups B1 to B4, phase information indicating the position of the block having the maximum similarity with the first block A is generated.

Further, when n is 3 or more, the phase adjusting unit 16 sets the ^{size of the reference source block in the original image LL 0} to 2 each in the horizontal and vertical directions, where j is an integer from 1 to n-1. ^In the first block multiplied by j and the (n-j) th-order low-frequency image LL ^n-j , the block having the same phase position as the reference block in the ^n-th-th order low-frequency image LL n and the block are horizontally and vertically. direction, a diagonal direction to a total displaced from each pixel to j pixel (j + 1) the degree of similarity calculated for the ^two second block group, only if the similarity exceeds the threshold value, the aligning section 12 The input alignment information may be output to the octave allocation unit 13.

Further, if the similarity obtained as described above is not high, it is considered that the accuracy of the alignment information is not high. Therefore, the phase adjusting unit 16 has a similarity with the first block of the second block group. The alignment information may be output to the octave allocation unit 13 only when the similarity of the maximum block exceeds the threshold value. In this case, the phase adjustment unit 16 determines the alignment information at the same time in addition to the phase adjustment.

The octave allocation unit 13 allocates the nth-order high-frequency component image to the 1-octave super-resolution high-frequency component according to the alignment information generated by the alignment unit 12, generates a 1-octave super-resolution high-frequency component image, and reconfigures the frequency. Output to unit 14. Further, according to the phase information generated by the phase adjusting unit 16, the (n−k) order high frequency component image is assigned to the (k + 1) octave super-resolution high-frequency component for k excluding 0, and the (k + 1) octave super-resolution high frequency is assigned. A component image is generated and output to the frequency reconstruction unit 14.

Similar to the first embodiment, the frequency reconstruction unit 14 uses the original image as a low frequency component and the (k + 1) octave super-resolution high-frequency component image allocated by the octave allocation unit 13 as a high-frequency component, and uses the n-th order frequency. Reconstruction processing is performed to generate an n-octave super-resolution image, which is output to the outside.

A computer can be preferably used to function as the above-mentioned super-resolution device 3, and such a computer provides a program describing processing contents for realizing each function of the super-resolution device 3 of the computer. This can be realized by storing the program in a storage unit and reading and executing this program by the CPU of the computer. This program can be recorded on a computer-readable recording medium.

As described above, the invention according to the third embodiment describes the reference source block in the first block whose size is doubled in the horizontal and vertical directions, and in the (n-1) phase low frequency image LL ^n-1 . The similarity is calculated for the second block having the same phase position as the referenced block and the second block group consisting of the blocks whose phase is shifted from the second block, and the similarity with the first block in the second block group. Includes in the alignment information the phase information indicating the position of the block where is maximum. As a result, the accuracy of the alignment information can be increased, so that it is possible to perform super-resolution enlargement of n octaves with high accuracy and high accuracy.

Although the above embodiments have been described as typical examples, it will be apparent to those skilled in the art that many modifications and substitutions can be made within the spirit and scope of the present invention. Therefore, the present invention should not be construed as being limited by the above-described embodiments, and various modifications and modifications can be made without departing from the scope of claims. For example, it is possible to combine a plurality of constituent blocks described in the configuration diagram of the embodiment into one, or to divide one constituent block.

The program may also be recorded on a computer-readable medium. It can be installed on a computer using a computer-readable medium. Here, the computer-readable medium on which the program is recorded may be a non-transient recording medium. The non-transient recording medium is not particularly limited, but may be, for example, a recording medium such as a CD-ROM or a DVD-ROM.

1, 2, 3 Super-resolution device 11 Frequency resolution unit 12 Alignment unit 13 Octave allocation unit 14 Frequency reconstruction unit 15 Similarity determination unit 16 Phase adjustment unit

Claims (4)

It is a super-resolution device that generates a super-resolution image obtained by enlarging the original image by n octaves.
A frequency decomposition unit that generates a frequency decomposition image by performing nth-order frequency decomposition processing on the original image,
An alignment unit that generates alignment information indicating the positional relationship between the reference source block in the original image and the reference destination block similar to the reference source block in the n-th order low frequency component image generated by the frequency decomposition unit. When,
Let k be an integer from 0 to n-1, and according to the alignment information, the (nk) order high frequency component image generated by the frequency decomposition unit is assigned to the (k + 1) octave super-resolution high frequency component (k + 1). ) Octave super-resolution Octave allocation part that generates high-frequency component images,
About the first block whose size is doubled in each of the horizontal and vertical directions for the reference source block, and the second block having the same phase position as the reference ^{block in the (n-1) order low frequency image LL n-1.} A similarity determination unit that calculates the similarity and outputs the alignment information to the octave allocation unit only when the similarity exceeds the threshold value.
A frequency reconstruction unit that generates a super-resolution image by performing n-th order frequency reconstruction processing using the original image as a low-frequency component and the (k + 1) octave super-resolution high-frequency component image as a high-frequency component.
A super-resolution device characterized by being equipped with. It is a super-resolution device that generates a super-resolution image obtained by enlarging the original image by n octaves.
A frequency decomposition unit that generates a frequency decomposition image by performing nth-order frequency decomposition processing on the original image,
An alignment unit that generates alignment information indicating the positional relationship between the reference source block in the original image and the reference destination block similar to the reference source block in the n-th order low frequency component image generated by the frequency decomposition unit. When,
Let k be an integer from 0 to n-1, and according to the alignment information, the (nk) order high frequency component image generated by the frequency decomposition unit is assigned to the (k + 1) octave super-resolution high frequency component (k + 1). ) Octave super-resolution Octave allocation part that generates high-frequency component images,
The first block whose size is doubled in the horizontal and vertical directions for the reference source block, the second block having the same phase position as the reference ^{block in the (n-1) order low frequency image LL n-1, and the reference block.} The similarity is calculated for the second block group consisting of blocks whose phases are shifted in the second block, and the phase information indicating the position of the block having the maximum similarity with the first block in the second block group is obtained. The phase adjustment unit included in the alignment information and
A frequency reconstruction unit that generates a super-resolution image by performing n-th order frequency reconstruction processing using the original image as a low-frequency component and the (k + 1) octave super-resolution high-frequency component image as a high-frequency component.
Super resolution device you wherein it comprises a. The super-solution according to claim 2 , wherein the phase adjusting unit outputs the alignment information to the octave allocation unit only when the similarity of the block having the maximum similarity exceeds the threshold value. Image device. A program for operating a computer as the super-resolution device according to any one of claims 1 to 3.

Images