JP2008159200A - Hologram reproducing apparatus, hologram reproducing method, reader and read method - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To achieve robust reproduction in a hologram reproducing apparatus. <P>SOLUTION: Two-dimensional FFT based on the output image of an image sensor is performed, and respective clock signals in an X direction and a Y direction included in the images are extracted from the analyzed result. Since cycle lines corresponding to a data pixel cycle in the X direction and the Y direction in reproducing images respectively appear in the clock signals, a resembling position is specified by obtaining the intersection. Since the resembling position can be specified using the whole including not only sync in the image but also user data, more robust reproduction is realized. <P>COPYRIGHT: (C)2008,JPO&INPIT

Description

The present invention reproduces a hologram recording medium in which data is recorded for each hologram page by interference fringes between reference light and signal light generated by modulation by a two-dimensional spatial light modulator according to recording data. The present invention relates to a hologram reproducing apparatus and a method thereof.
The present invention also relates to a reading apparatus and method for reading the value of each pixel in the two-dimensional image on a medium on which data is recorded by a two-dimensional image in which data bit values are stored in units of required pixels.

In the hologram recording / reproducing system, particularly in the hologram recording / reproducing system in the field of optical storage systems, for example, an SLM (spatial light modulator) such as a transmissive liquid crystal panel or DMD (Digital Micromirror Device) is used as the light intensity modulation. The intensity modulation is performed so that a pattern arrangement of bit 1 (for example, light intensity = strong) and bit 0 (for example, light intensity = weak) is obtained.
At this time, in the SLM, for example, as shown in FIG. 2, a signal light is generated at the central portion according to the recording data according to the recording data to generate signal light, and the light is transmitted in a ring shape around the reference light. It is designed to produce light. Then, the signal light modulated in accordance with the recording data is irradiated onto the hologram recording medium together with the reference light, whereby the interference fringes between the signal light and the reference light are recorded as data on the hologram recording medium. .

  When reproducing data, the SLM generates only the reference light and irradiates the hologram recording medium with the reference light to obtain diffracted light corresponding to the interference fringes. An image corresponding to this diffracted light is imaged on an image sensor such as a CCD (Charge Coupled Device) sensor or a CMOS (Complementary Metal Oxide Semiconductor) sensor, and the amplitude value of each data pixel (that is, each pixel of the SLM) is obtained. . Then, reproduction data is obtained based on the amplitude value of each data pixel.

  The hologram recording / reproducing method for irradiating the signal light and the reference light on the same optical axis in this way is known as a coaxial method. For example, such a coaxial method is generally used in hologram recording / reproducing. Due to problems such as optical distortion and magnification, it is practically difficult to match each pixel of the SLM and each pixel of the image sensor exactly one to one. That is, it is practically difficult to make the reproduction light corresponding to each pixel of the SLM strictly enter each assumed pixel on the image sensor.

Therefore, at the time of recording, when arranging data in the signal light, predetermined pattern data called a sync is inserted at predetermined intervals, and at the time of reproduction, alignment based on the position of the sync is performed, and then the amplitude of each pixel is set. The value is calculated.
By inserting the sink in this way and matching the position where the sync pattern was actually detected during playback, the position from the ideal incident position of the playback light due to optical distortion, etc. It is possible to correct the deviation so that an appropriate read operation is performed.

  However, the above-described shift in the incident position does not always occur in units of one pixel, and it is expected that the shift occurs in units of less than pixels. When a shift of less than a pixel unit occurs as described above, it is impossible to properly detect the sync for performing the alignment as described above, and not only the data reading but also the alignment remains.

Therefore, for example, the number of pixels of the image sensor is set to n times the number of pixels of the SLM (for example, 2 × 2 = 4 times, but not necessarily an integral multiple) so as to cope with such a shift in units of less than pixels. The resolution on the image sensor side with respect to the reproduction light for one pixel of the SLM is increased.
For example, in the conventional general method, as 2 × 2 oversampling, reproduction light for one pixel of the SLM is received by 2 × 2 = 4 pixels on the image sensor. As a result, it is possible to obtain a resolution four times that of one pixel of the SLM.

  Further, conventionally, in order to further improve the accuracy, the respective values obtained by the 2 × 2 oversampling are interpolated in this way, and further 2 × 2 up-conversion is performed ( Up-conversion to 4x4). That is, by performing up-conversion to 4 × 4, the resolution of one pixel of the SLM can be increased 16 times.

  For example, the readout position of each pixel can be corrected in units of less than one pixel by such a method, and the pixel amplitude value can be appropriately calculated.

Here, a specific example of the conventional hologram recording / reproducing method, which has been outlined above, will be described.
In the explanation, the following terms are defined.
・ Symbol ・ ・ ・ ・ Recording encoding minimum unit: As a specific example, one byte (= 8 bits) of recording data is converted into a square data unit of 4 bits × 4 bits.・ Data unit of 6 symbols × 6 symbols (24 bits × 24 bits) ・ Page: Unit of the total amount of data finally spread in the signal light area

First, a conventional recording data format will be described with reference to FIGS.
FIG. 47 schematically shows a modulation pattern in the SLM. FIG. 48 shows an example of a sync pattern inserted in the recording data.
First, in FIG. 47, conventionally, when forming data as one page, data is laid in a circular signal light area with one subpage as a minimum unit as shown in the figure. .

  When data is laid down, syncs are inserted at predetermined intervals. Conventionally, a page sync as shown in the figure is inserted at the top of each page (in this case, the position at the left end of the top row).

For page sync, an area for one subpage is allocated as shown in FIG. As the data pattern, as shown in the figure, all the bits of 2 × 2 = 4 symbols located in the center of one subpage and 4 bits × 4 bits (that is, one symbol) at the center are expressed as “ It is assumed that a pattern in which all other bits are set to “0” is defined as “1”. In other words, in this case, the data pattern of the central four symbols is as follows:
"00000000
00000000
00111100
00111100
00111100
00111100
00000000
00000000 "
It becomes. All bits other than the central four symbols are “0”.

Further, in each page, a subpage sync as shown in FIG. 48B is attached to each subpage which is the minimum unit of data padding as described above.
As this sub-page sync, the same pattern as that of the above-mentioned center 4 symbols of the page sync is inserted into the position of the center 4 symbols in one sub-page.

In this way, in the conventional format, data is laid out in the signal light area with the minimum laying unit as one subpage, and for sync, a page sync for each page (for one subpage), and A subpage sync (for 4 symbols) is inserted for each subpage in one page.
According to such a format, conventionally, as shown in FIG. 47, user data (sink) corresponding to 3552 symbols as the number of effective symbols in the signal light area of radius = 154 pixels (pixel = one pixel of SLM). Other data) can be packed. That is, the effective capacity for one page is 3552 symbols.

FIG. 49 schematically shows an example of a conventional hologram recording / reproducing method performed based on such a conventional recording data format. In FIG. 49, the recording / playback processing procedure according to the conventional method is shown in blocks.
First, at the time of recording, recording modulation encoding is performed on the recording data as shown in <1>. In this case, for example, a so-called sparse code is used as the recording modulation code. Specifically, in this sparse code, 1 byte (= 8 bits) of the recording data is converted into a 4 × 4 bit square data array (that is, the above-described 1 symbol), and among the 16 bits after conversion, Only 3 bits are “1” and the remaining 13 bits are “0”.

Then, after recording data is divided into symbols by recording encoding such as sparse encoding, mapping to subpages is performed as shown in <2>. That is, each symbol obtained by encoding is mapped in a subpage.
At this time, as shown in FIG. 47, a subpage sync is inserted into each subpage at the 4 symbols in the center. Therefore, in this <2> mapping process, the generated symbols for one page (in this case, 3552 symbols) are grouped into 32 symbols, and each of these 32 symbols is arranged in order, and the four symbols at the center are shown in FIG. Each subpage is mapped so that subpage syncs according to a predetermined data pattern as shown in FIG. 48 are arranged, and a total of 111 subpages are generated.

When each subpage is generated, mapping to the entire page is performed as shown by <3>. That is, one page sync as shown in FIG. 48A and 111 generated subpages are mapped (spread) in the signal light area according to the format.
At this time, since the SLM also generates reference light, the data pattern for all effective pixels of the SLM includes the pattern of the reference light area (predetermined pixels) together with the data pattern for one page thus generated. Only “1”), a gap area (a gap area between the signal light area and the reference light area) (all “0”), and a pattern outside the reference light area (all also “0”). A data pattern combined with is generated.
By driving and controlling the SLM based on this data pattern, data is recorded on the hologram recording medium by interference fringes between the signal light and the reference light.

  Note that the mapping order of the symbols <2> to the sub-pages and the mapping order of the sub-pages <3> to the entire page may be any order. It is common practice to go and go down one step and return to the left end until the right end of the bottom step.

And the reproduction | regeneration about the data recorded on the hologram recording medium in this way is performed as follows.
First, as shown by <4>, 2 × 2 oversampling is performed. That is, as described above, the number of pixels of the image sensor with respect to the number of pixels of the SLM is set to be, for example, 2 × 2 = 4 times, and an image for one pixel of the SLM is displayed on the image sensor. 2 × 2 = light is received by four pixels.
In addition, in this <4> 2 × 2 oversampling, in order to improve the frequency characteristics of the read signal obtained by oversampling, some filter such as a low-pass filter (LPF) is actually passed. Yes.

In <5>, the readout signal oversampled at 2 × 2 as described above is further up-converted to 4 × 4.
In other words, 2 × 2 up-conversion is performed by interpolating the respective values obtained by the above 2 × 2 oversampling (and filtering) (up-conversion to 4 × 4).

After these oversampling and up-conversion, a process for appropriately calculating the amplitude value of each pixel is performed in consideration of the optical distortion as described above.
First, page sync detection is performed as shown in <6>. That is, the position of the page sync that should be in the read signal obtained by the oversampling / up-conversion is detected based on the data pattern of the page sync defined in advance as described in FIG. is there.

Here, in each pixel of the image sensor, for example, the amplitude value is obtained with 256 gradations of 0 to 255. That is, each value obtained through the oversampling / up-conversion as described above is a value represented by 0 to 255.
As a specific method for detecting a page sync, the average value of the data pattern is set to 0 by adding a certain value (negative number) to the data pattern (0, 1 pattern) of the page sync known in advance. Generate a sync reference pattern, perform correlation calculation with the sync reference pattern around the position where the page sync should be in the signal obtained by oversampling / up-conversion (this can also be estimated from the format), As a result, a method of specifying the position where the correlation value becomes the largest as the sync position is adopted (such a position determination method is called template matching by correlation calculation).
By specifying the position of the page sync in this way, first, it is possible to roughly align the entire page.

After that, as shown in <7>, sub-page sync is detected. That is, the position of each subpage sync in the page is specified.
If a certain reference point in the page as a page sync is determined as described above, the position where the target sub-page sync should be located is how many pixels away from the determined reference point based on the recording format. Can be identified. Therefore, with the detected page sync position as a reference, the sub-page sync is detected by template matching by correlation calculation in the same way as in the previous page sync detection, around the position where the target sub-page should be estimated according to the format. Identify the location.

When each subpage sync is detected, a process of resampling to 1 × 1 is performed as shown in <8>. That is, based on the detected position of each subpage sync, the position of the readout signal corresponding to each data pixel of the SLM (also simply referred to as the pixel position) is identified from the oversampled and upconverted signals, and the amplitude thereof Get the value.
Specifically, the position of each pixel is specified for each subpage based on the detected position of each subpage sync, and the amplitude value of each pixel is acquired. At this time, in order to identify data in symbol units corresponding to the sparse encoding method described below, the acquired pixel values are collected in symbol units here.

In the next <9>, data identification is performed in symbol units. That is, data identification is performed on the basis of the values collected in symbol units as described above to detect the bit value of each pixel.
Here, in the case of a sparse code in which 3 bits out of 1 symbol = 16 bits employed in this case are “1”, “sort detection” in which three pixels having a large value are set to bit “1”. A so-called data identification method can be applied. Therefore, for each symbol, data identification is performed by such sort detection, and the final value of the recording bit is detected.

Finally, the recording code (sparse code) is decoded as shown in <10>.
That is, each bit value obtained for each symbol by data identification in the symbol unit is rearranged according to the order mapped at the time of recording, and further, the rearranged bit value is decoded (demodulation of sparse code). Returning the symbol = 16 bits → 1 byte = 8 bits, reproduction data having the same content as the recording data is obtained.

  As described above, data recording to the hologram recording medium and reproduction of the recorded data are performed.

In addition, about the related prior art, the following patent documents can be mentioned.
JP 2006-196044 A Japanese Patent Laid-Open No. 2001-14418

However, the conventional hologram recording / reproducing method as described above has the following problems.
First, the conventional hologram recording / reproducing method promotes an increase in the number of syncs and an increase in the sync size, and there is a problem that the capacity of user data is reduced correspondingly.
That is, in the above-described conventional method, since a sync (subpage sync) is used to specify the position of each data pixel, many syncs are required to improve detection accuracy, and the sync size is increased. There is a need. In this sense, there is a problem in that the number of syncs increases and the sync size increases, and the capacity of user data that can be recorded in one page must be reduced.

Further, the above-described conventional method has a problem in that only the sync is used to detect the position of each data pixel, and user data other than the sync is not used effectively.
In other words, since only the sync is used to specify the position of each data pixel in this way, the influence when the sync is damaged is large, and if the sub-page sync is damaged and cannot be detected. May immediately lose data in the subpage.
In addition, if the page sync is damaged and cannot be detected, the entire page data may be lost.

  From these points, the conventional method is not robust against a decrease in SNR (S / N ratio) of the detected image. Further, it is difficult to improve the recording density by reducing the number of syncs and reducing the sync size.

In the first place, the above-mentioned conventional method also has a drawback in that the sink is used in a form different from the original usage.
That is, in the conventional storage apparatus and communication apparatus, the “synchronization” is for frame synchronization after bit synchronization is taken, and not for bit synchronization. That is, it can be said that the concept of using the sync for specifying the position of each data pixel corresponding to the bit synchronization as in the conventional method described above is incorrect in the usage of the sync.

Another point is that the image is not robust against various fluctuations, distortions, and deterioration of the image.
For example, as a form of distortion of the reproduced image, it can cope with movement in the plane direction, but is not robust when the image is enlarged or reduced.
Also, it is not robust against the rotation of the playback image.

By the way, in the above description, it has been pointed out that the conventional method uses a sync for bit synchronization, and the use of a sync different from the original is problematic. In the field of recording and reproduction, a concept has been proposed in which a sync for bit synchronization is not used and the position of each data pixel is performed after extracting a clock signal based on user data.
For example, Patent Document 1 mentioned above corresponds to this.

FIG. 50 schematically shows a hologram recording / reproducing method based on the invention described in Patent Document 1.
In this method, one-dimensional image data is subjected to Fourier transform processing such as FFT on a value read by a one-dimensional image sensor (in this case, oversampling or up-conversion may be performed) to construct an image. Decompose each frequency element.
As a result of performing the Fourier transform in this way, the frequency component corresponding to the pixel period of the spatial light modulator on the recording side becomes relatively strong, so a signal in the vicinity of that frequency is extracted, and this is referred to as a clock signal. can do. That is, a clock signal for specifying the position of each recording pixel as shown in FIG. 50B can be extracted.

However, this is a technique based on a one-dimensional image, and cannot be applied to hologram recording / reproduction using a two-dimensional image as it is.
For example, if it is applied to a two-dimensional image, it is conceivable that the same Fourier transform and clock signal component are extracted in line order in the horizontal direction (X direction). For example, as shown in FIG. Thus, when the reconstructed image is tilted with respect to the image sensor by rotation of the reconstructed image, the data pixel such as a portion ■ (that is, bit “0”) increases in a certain line as shown in the figure. In some cases, it may become impossible to extract a clock signal with a period corresponding to the number.
Alternatively, even if the reproduced image does not rotate, depending on the line in the horizontal direction (X direction), it matches the vicinity of the boundary in the Y direction of the spatial light modulator, and the gray scale value is properly captured by the image sensor. In this case, the clock signal may not be extracted successfully.
In the first place, since the clock signal in such a one-dimensional image only represents periodicity in one direction, even if this is applied to a two-dimensional image as it is, an appropriate sample position is specified. Obviously you can't.
Thus, only by the one-dimensional concept as described in Patent Document 1, it is impossible to properly perform bit synchronization (that is, specify the position in units of data pixels) for a two-dimensional image.

In view of the above problems, in the present invention, the hologram reproducing apparatus is configured as follows.
That is, the hologram reproducing apparatus of the present invention records data for each hologram page unit by interference fringes between the reference light and the signal light generated by being modulated by the two-dimensional spatial light modulator according to the recording data. A hologram reproducing apparatus that reproduces the hologram recording medium is provided with reference light irradiation means for irradiating the hologram recording medium with the reference light.
Further, an image acquisition unit is provided for obtaining two-dimensional image data based on a result of receiving diffracted light corresponding to the hologram page unit data obtained by irradiating the hologram recording medium with the reference light.
Moreover, an analysis unit is provided that performs two-dimensional Fourier transform processing based on the two-dimensional image data and performs frequency analysis on the plane wave element included in the two-dimensional image data.
Further, a peak portion of the power spectrum is searched for within the first predetermined range and the second predetermined range for the analysis result by the analyzing means, and the peak portion detected within the first predetermined range is detected. The first direction clock information indicating the period, phase and normal direction of the plane wave specified based on the first phase, and the period, phase and normal of the plane wave specified based on the peak portion detected within the second predetermined range. Clock information acquisition means for acquiring second direction clock information indicating the direction is provided.
In addition, pixel position specifying means for specifying the position of the spatial light modulator in units of pixels in the two-dimensional image data based on the first direction clock information and the second direction clock information.
Furthermore, an amplitude value acquisition unit that acquires an amplitude value of each pixel based on the position information specified by the pixel position specifying unit is provided.

In the present invention, the reading apparatus is configured as follows.
That is, the reading device of the present invention is a reading device for reading the value of each pixel in the two-dimensional image with respect to a medium on which data is recorded by a two-dimensional image in which data bit values are stored in units of required pixels. First, light irradiation means for performing light irradiation on the image on the medium is provided.
Moreover, the image acquisition means which acquires the two-dimensional image data according to the said two-dimensional image on the said medium based on the result of having detected the light obtained according to the light irradiation to the said medium is provided.
In addition, a non-linear processing unit that performs non-linear processing on the two-dimensional image data obtained by the image acquisition unit is provided.
Further, an analysis unit is provided that performs a two-dimensional Fourier transform process on the two-dimensional image data processed by the nonlinear processing unit and performs a frequency analysis on a plane wave element included in the two-dimensional image data.
Further, a peak portion of the power spectrum is searched for within the first predetermined range and the second predetermined range with respect to the analysis result by the analyzing means, and the peak detected within the first predetermined range is detected. First direction clock information indicating the period, phase and normal direction of the plane wave specified based on the portion; and the period, phase and phase of the plane wave specified based on the peak portion detected within the second predetermined range. Clock information acquisition means for acquiring second direction clock information representing the line direction is provided.
In addition, a pixel position specifying unit is provided for specifying a position in units of pixels in the two-dimensional image recorded on the medium in the two-dimensional image data based on the first direction clock information and the second direction clock information. .
Furthermore, an amplitude value acquisition unit that acquires an amplitude value of each pixel based on the position information specified by the pixel position specifying unit is provided.

Each plane wave element constituting the two-dimensional image data is analyzed by performing a two-dimensional Fourier transform process based on the two-dimensional image data obtained according to the image recorded on the medium as described above. be able to.
In the analysis result by the two-dimensional Fourier transform, a plane wave (specified by the period, phase, and direction of the wave) as a clock signal in the first direction (for example, the X direction) included in the two-dimensional image data. A plane wave component as a clock signal in the second direction (for example, the Y direction) appears as a peak portion of the power spectrum in the first predetermined range and the second predetermined range, respectively. Therefore, if the peak portions of the power spectrum are searched for in the first predetermined range and the second predetermined range as described above, the X-direction clock information and the peak information are detected based on the detected peak portion components. Y-direction clock information can be obtained.
In this way, if X-direction clock information and Y-direction clock information representing information about the period, phase, and direction of each clock signal in the X direction and the Y direction can be obtained, they are recorded on the medium based on them. The position of each image in pixel units can be specified, and the amplitude value of each pixel can be obtained based on the position information.

  In the reading apparatus / reading method of the present invention, two-dimensional Fourier transform processing is performed on the two-dimensional image data that has undergone nonlinear processing. Thus, two-dimensional Fourier transform is performed on the image after nonlinear processing. As a result of applying, a strong peak appears in a portion corresponding to the clock signal as shown in FIG. In other words, if this is done, it is not determined how many pixels on the light receiving element side will receive the image of one pixel of the image recorded on the medium in advance, and the approximate frequency of the clock signal cannot be estimated. Even when the search range cannot be limited to a narrow range, the respective clock information can be appropriately acquired by searching for peak components in a predetermined range such as the vicinity of the X axis or the vicinity of the Y axis, for example. .

  As described above, in the present invention, the X-direction clock information (X-direction) is obtained from the frequency analysis result by the two-dimensional Fourier transform process based on the two-dimensional image data obtained according to the two-dimensional image recorded on the medium. And the Y-direction clock information (information on the wave period, phase and direction of the clock component in the Y direction). Based on the X-direction clock information and the Y-direction clock information, the position of each pixel in the image recorded on the medium can be specified, and the value of each pixel is acquired based on the position information. Can do.

  At this time, the X-direction clock information and the Y-direction clock information acquired from the analysis result include direction information for the X-direction clock component and the Y-direction clock component together with the wave period and phase. Yes. From this, by specifying each pixel position based on the X-direction clock information and the Y-direction clock information as described above, the position of each pixel is specified appropriately corresponding to the rotation of the image. be able to. Of course, each pixel position can also be specified appropriately corresponding to the enlargement / reduction of the image. Furthermore, for example, each pixel position can be specified appropriately corresponding to distortion in which the clock frequency is different between the X direction and the Y direction.

  Further, according to the present invention, the position of each pixel can be specified without performing alignment by sync as in the prior art. According to this, it is possible to eliminate the need for a sub-page sync for specifying the position of each pixel, which has been conventionally required, and to increase the recording capacity of user data accordingly. That is, it is possible to improve the recording density of user data.

  Further, according to the present invention, the position of each pixel can be specified using not only the sync portion in the image but also the entire image data including user data. That is, conventionally, when the subpage sync is damaged, there is a possibility that the position of each pixel is not properly specified in the area and the data in the area is completely destroyed. In addition, since the position of each pixel can be specified using the entire two-dimensional image data, more robust data reading can be performed in that sense.

  Also, if the amplitude value of each pixel can be obtained without specifying the position using the subpage sync as described above, the position where each bit value corresponds to the format is determined after bit synchronization. Therefore, it is possible to adopt a normal sync usage method in which frame synchronization is performed by a sync for performing the sync.

Hereinafter, the best mode for carrying out the invention (hereinafter referred to as an embodiment) will be described.
The description will be given in the following order.

1. 1. Configuration of playback device 2. Configuration of data reproduction unit 3. Re-sampling position specifying method as embodiment 3-1. X-direction and Y-direction differentiation 3-2. Nonlinear processing 3-3. Two-dimensional Fourier transform processing 3-4. Extraction of clock signal component 3-5. Phase shift processing 3-6. Inverse Fourier transform processing 3-7. Extraction of zero-cross line 3-8. Extraction of grid point of zero cross line 4. 4. Internal configuration of resampling position specifying unit Effects of the embodiment 6. Experimental results Modified example

In the following description, the term “clock signal” is used. Note that the “signal” in this case is a two-dimensional signal having a scalar quantity and is equivalent to a two-dimensional image. deep. In the coordinate system for representing an image, the X direction is the right direction in the figure, and the Y direction is the downward direction in the figure.

1. Configuration of playback device

FIG. 1 is a block diagram showing an internal configuration of a recording / reproducing apparatus (recording / reproducing apparatus 1) that can be configured based on the present invention. In FIG. 1, only the configuration of the optical system, the recording data modulation system, and the reproduction system of the recording / reproducing apparatus 1 is mainly extracted and the other parts are omitted.

First, in this embodiment, a so-called coaxial system is adopted as a hologram recording / reproducing system. That is, the signal light and the reference light are arranged on the same axis, and both of them are applied to the hologram recording medium 5 to record data by interference fringes, and only the reference light is applied to the hologram recording medium 5 during reproduction. By doing so, the data recorded by the interference fringes is reproduced.
Further, in this case, the hologram recording medium 5 in the figure is a so-called reflection type hologram recording medium having a reflection film, and the recording / reproducing apparatus 1 has a configuration corresponding to such a reflection type hologram recording medium 5. It is done.

  In FIG. 1, the recording / reproducing apparatus 1 is first provided with an optical system including a laser diode LD, a collimator lens 2, an SLM (spatial light modulator) 3, a beam splitter 4, an objective lens OL, and an image sensor 6. .

The laser diode LD is provided as a light source for obtaining laser light for recording and reproduction. Light emitted from the laser diode LD passes through the collimator lens 2 and is converted into parallel light, which is guided to the SLM 3. For example, a transmissive liquid crystal panel is used as the SLM 3.
The light subjected to spatial light modulation by the SLM 3 passes through the beam splitter 4 and is guided to the objective lens OL side. Then, the light passes through the objective lens OL and is irradiated onto the hologram recording medium 5 set at a predetermined position.

  At the time of recording, spatial light modulation according to the recording data is performed in the SLM 3 as will be described later, and the parallel light thus modulated passes through the objective lens OL to be converged light, and the hologram recording is performed. The light is condensed on the medium 5.

  During reproduction, the hologram recording medium 5 is irradiated with light from the laser diode LD after being modulated by the SLM 3 through the same path as described above, thereby recording data as described later. The diffracted light according to is obtained. The diffracted light is converted into parallel light as reflected light from the hologram recording medium 5 through the objective lens OL, and then reflected by the beam splitter 4 and guided to the image sensor 6 side. The image sensor 6 is, for example, a CCD (Charge Coupled Device) sensor or a CMOS (Complementary Metal Oxide Semiconductor) sensor, and receives reflected light (diffracted light) from the hologram recording medium 5 guided as described above. , Convert to electrical signal.

  Also in the case of the present embodiment, it is assumed that oversampling and up-conversion similar to the conventional one is performed, and an image of one pixel of the SLM 3 is represented by n pixels on the image sensor 6 (n> 1: n is an integer). But not necessarily). Also in this case, for example, it is assumed that an image of one pixel of the SLM 3 is received by four pixels (2 pixels × 2 pixels) on the image sensor 6.

Here, with reference to FIG. 2 and FIG. 3, a method for recording and reproducing data on the hologram recording medium 5 by the optical system described above will be described. FIG. 2 shows the recording method, and FIG. 3 shows the reproduction method.
In FIG. 2, only the SLM 3 and the objective lens OL are extracted from the optical system shown in FIG. 3A similarly shows only the SLM 3 and the objective lens OL, and FIG. 3B shows only the objective lens OL and the image sensor 6 extracted.

First, at the time of recording shown in FIG. 2, the SLM 3 performs light (in which a data array of “0” and “1” is formed based on the reference light and the recording data with respect to the incident light from the collimator lens 2 ( (Hereinafter referred to as “signal light”) is intensity-modulated so as to be arranged concentrically.
The intensity-modulated light (that is, the reference light and the signal light) is collected on the hologram recording medium 5 by the objective lens OL, and the interference fringes of the reference light and the signal light formed thereby are used as data for the hologram recording medium 5. To be recorded.

  At the time of reproduction, first, as shown in FIG. 3A, the incident light from the collimator lens 2 is intensity-modulated so that only the reference light pattern is output by the SLM 3 and collected on the hologram recording medium 5. Shine. At this time, the condensed light is diffracted by interference fringes corresponding to the data pattern recorded on the hologram recording medium 5 and output as reflected light from the hologram recording medium 5. That is, the diffracted light has an intensity modulation pattern reflecting recorded data as shown in the figure, and data reproduction is performed based on the result of reading the intensity modulation pattern of the diffracted light with the image sensor 6. To be.

Here, in the SLM 3, as described above, the reference light and the signal light are generated corresponding to the recording / reproducing time. Therefore, in the SLM 3, a reference light area A1, a signal light area A2, and a gap area A3 as shown in FIG. 4 are defined. That is, as shown in the figure, a predetermined circular area including the central portion of the SLM 3 is defined as the signal light area A2. An annular reference light area A1 that is concentric with the signal light area A2 is defined with respect to the outer peripheral portion with a gap area A3 therebetween.
The gap area A3 is defined as a region for avoiding the reference light at the time of reading leaking into the signal light area A2 and becoming noise.

At the time of recording, a predetermined pixel in the reference light area A1 is set to “1” (light intensity = high), and other pixels are set to “0” (light intensity = weak), and the gap area A3 and the reference light area By setting all the outer peripheral parts from A1 to “0” and making each pixel in the signal light area A2 a required “0” “1” pattern according to the recording data, as shown in FIG. It is possible to generate and output such reference light and signal light.
Further, at the time of reproduction, only the reference light area A1 is set to the same pattern of “0” and “1” as at the time of recording, and all other areas are set to bit “0”, as shown in FIG. Only reference light can be generated and output.

In addition to the optical system described so far, the recording / reproducing apparatus 1 shown in FIG. 1 includes a data modulation unit 7 for realizing the modulation pattern in the SLM 3 as described above.
The data modulation unit 7 receives recording data supplied at the time of recording, and generates a data pattern in the signal light area A2 so as to cover the recording data in the signal light area A2 according to a predetermined format.
Further, a data pattern is generated in which the reference light area A1 is a predetermined “0” “1” pattern, and the outer peripheral portion of the gap area A3 and the reference light area A1 is all bits “0”. This data pattern and the data pattern in the signal light area A2 are combined to generate a data pattern for all effective pixels of the SLM 3.
Based on this data pattern, each pixel of the SLM 3 is driven and controlled, whereby modulated light (signal light and reference light) during recording as shown in FIG. 2 can be obtained.

  On the other hand, during reproduction, only the reference light area A1 has a pattern of “0” and “1” that is the same as that during recording, and other areas all generate “0”, and each pixel of the SLM 3 is generated based on this data pattern. By controlling the drive, it is possible to obtain the modulated light during reproduction as shown in FIG.

Also in the case of the present embodiment, the data modulation section 7 sparsely encodes the input recording data and converts it into symbol units (4 × 4 = 16-bit square units) for each 8-byte data. Then, it is assumed that a data pattern in which these symbols are spread in the signal light area A2 according to a predetermined format is generated.
Also in this case, sparse coding is performed so that, for example, 3 bits out of 1 symbol = 16 bits are “1” and the remaining 13 bits are all “0”, as in the conventional case. Suppose that

Further, the recording / reproducing apparatus 1 is provided with a data reproducing unit 8 for reproducing recorded data based on a detection signal (detection value) of each pixel in the image sensor 6.
The data reproducing unit 8 performs each process to be described later, specifies the position of each data pixel (each pixel of the SLM) in the detection signal of the image sensor 6, and detects the amplitude value of each data pixel based on the position. Yes (resampling). After that, the recording data is reproduced by performing data identification based on the amplitude value of each data pixel and decoding processing of the sparse code.

2. Data playback unit configuration

FIG. 5 shows an internal configuration of the data reproducing unit 8 shown in FIG.
In the data reproduction unit 8, an LPF (low-pass filter) 20, an up-conversion unit 21, a resampling position specifying unit 22, a resampling unit 23, a page sync alignment unit 24, a symbol extraction unit 25, as shown in FIG. A data identification unit 26 and a sparse code decoding unit 27 are provided.
The detection output (sensor output in the figure) from the image sensor 6 shown in FIG. 1 is input to the LPF 20 in the data reproduction unit 8, where predetermined filtering processing for noise removal and frequency characteristic improvement is performed. Then, the data is input to the up-conversion unit 21.

As described above, even in the case of this embodiment, for example, as 2 × 2 oversampling, the image sensor 6 receives an image of one pixel of the SLM 3 with 2 × 2 = 4 pixels. Has been.
After the 2 × 2 oversampling in this way, the up-conversion unit 21 receives the sensor output after the filter processing by the LPF 20 and performs up-conversion processing at a predetermined magnification. Also in this case, as the up-conversion process, for example, a magnification of 2 is set, and thereby up-conversion to 4 × 4 is performed. That is, as a result, a resolution four times that of the SLM 3 is obtained. In terms of the number of pixels, 4 × 4 = 16 pixel signals are obtained for one pixel of the SLM 3.

The resampling position specifying unit 22 receives the signal (image data) processed by the up-conversion unit 21 and performs processing for specifying the position of each pixel (data pixel unit) of the SLM 3 in the image data. Execute.
The resampling position specifying operation by the resampling position specifying unit 22 and the internal configuration for realizing it will be described later.

The re-sampling unit 23 receives information on the position of each pixel obtained as a result of the processing by the re-sampling position specifying unit 22 and the two-dimensional image data after the up-conversion processing by the up-conversion unit 21, and the two-dimensional image The amplitude value of the position specified by the position information of each pixel in the data is acquired (this is called 1 × 1 resampling).
As such resampling processing after each pixel position is obtained, two-dimensional signal interpolation may be performed based on the sampling theorem. Alternatively, interpolation processing generally performed in the field of conventional image processing may be performed. For example,
・ Nearest neighbor method
・ Bi-linear interpolation method
・ Cubic convolution method
・ Bicubic spline method
This can be performed by any of the interpolation processes.
Among these methods, for example, the nearest neighbor method selects the readout signal having the closest timing as the amplitude value of the pixel, and is effective when the oversampling rate is large. This nearest neighbor method has an advantage that the processing time can be shortened in that calculation processing based on a function or the like is unnecessary.
The cubic interpolation method is a piecewise cubic polynomial approximation of the function sin (x) / x used when interpolating based on the sampling theorem, and the processing load is larger than the nearest neighbor method. There is an advantage that highly accurate results can be obtained. In the case of the present embodiment, this cubic interpolation method (also simply referred to as Cubic interpolation) is employed for the resampling process.

By the resampling process of the resampling unit 23, an amplitude value at each position in the data pixel unit in the two-dimensional image data is obtained. That is, two-dimensional image data having the same rate (1 × 1 rate) as the pixels of the SLM 3 is obtained.
The page sync alignment unit 24 performs alignment processing on the basis of information on the amplitude value of each pixel in such two-dimensional image data and information on a predetermined data pattern as a page sync predetermined in the format. . That is, the position of each specified pixel on the format (that is, the position in the hologram page defined by the format) is specified.

The alignment processing by this page sync can be performed by a method similar to the conventional method. That is, based on the predetermined data pattern as the page sync as described above, the template matching is performed by the correlation calculation described above. Specifically, since the range where the page sync exists can be estimated to some extent based on the format, the correlation value with the data pattern is calculated within the estimated range, and the location giving the maximum To do.
As described above, unlike the conventional method, the alignment processing in the case of the present embodiment is an image whose amplitude value has already been obtained in units of data pixels (that is, a rate of 1 × 1) as described above. Therefore, the sink can be used in its original usage. The calculation amount of the alignment process can be significantly reduced as compared with the conventional case where the image after the 4 × 4 up-conversion is performed.

  The symbol extraction unit 25 extracts each symbol in the hologram page by using the position information on the format of each pixel specified by the page sync alignment unit 24. Then, the amplitude values of the respective pixels are collected for each extracted symbol and supplied to the data identification unit 26.

The data identification unit 26 performs data identification for each symbol using the amplitude value of each pixel for each symbol unit supplied from the symbol extraction unit 25.
In this case, since sparse encoding at the time of recording is performed such that only 3 bits out of 1 symbol = 16 bits are “1”, a sort detection method corresponding to this is adopted. That is, data identification is performed by setting the data bit values of the top three pixel positions having large amplitude values for each symbol unit to “1”, and the data bit values of all the remaining pixel positions as “0”. The identification of the data bit value of “1” is detected in symbol units.

  The value of each data bit for each symbol unit detected by the data identification unit 26 is supplied to the sparse code decoding unit 27, where the sparse code is decoded for each symbol, thereby 1 symbol → 1 byte. = Conversion is performed so as to return to 8 bits. Thereby, the recorded data is reproduced.

Although illustration is omitted here, a configuration for performing basic preprocessing that is normally performed in image processing may be added to a necessary position in the data reproducing unit 8 as appropriate. For example, a signal processing unit for performing density unevenness removal by AGC (Automatic Gain Control), light level correction, dark level correction, and the like is added.

3. Resampling position identification method as an embodiment

Subsequently, the resampling position specifying operation as an embodiment will be described.
The resampling position of this embodiment is specified by the following procedure.

-X direction and Y direction differentiation (3-1.)
・ Nonlinear processing (3-2.)
・ Two-dimensional Fourier transform processing (3-3.)
Extraction of clock signal component (3-4.)
・ Phase shift processing (3-5.)
・ Inverse Fourier transform processing (3-6.)
・ Zero cross line extraction (3-7.)
・ Lattice point extraction of zero-cross line (3-8.)

3-1. X-direction and Y-direction differentiation

First, FIGS. 6 and 7 show a sensor output image obtained based on the detection result of the image sensor 6 and an image after up-conversion by the up-conversion unit 21. FIG.
The sensor output image shown in FIG. 6 shows an image after processing by the LPF 20.
6 and FIG. 7 and the subsequent figures that deal with images, it is difficult to grasp the detailed state of the whole image, so that the positions are common and only a part thereof is enlarged.

First, as a premise, in the case of the present embodiment, it is assumed that the image size of the sensor output image shown in FIG. 6 is 1024 × 1024 pixels. As described above, since the oversampling rate in this case is 2 × 2, the pixel of the image sensor 6 (also referred to as “detector pixel” for the data pixel of the SLM) has one pixel of the SLM 3. Corresponding at a rate of 2 × 2. That is, an image for one pixel of the SLM 3 is received by four pixels on the image sensor 6.
Although this ratio may be considered to be substantially constant, the alignment state (phase) changes variously and also varies within the image. Although it is almost constant, it should be considered that the oversampling rate also changes and there are fluctuations in the image. In addition, the sensor output image includes various fluctuations, distortions, and deterioration.

  Further, as can be seen from the image size of the sensor output image and the oversampling rate, it is assumed that the number of effective pixels of the SLM 3 in this case is 512 × 512 pixels.

Comparing the sensor output image shown in FIG. 6 with the image after up-conversion shown in FIG. 7, it can be seen that the resolution shown in FIG. 7 is higher than the image shown in FIG. In this case, since the up-conversion rate is 2 × 2, the image size of the image shown in FIG. 7 is 2048 × 2048 pixels, which is twice as long as 1024 × 1024 pixels of the image shown in FIG.
In addition, it can be seen that the image size of 2048 × 2048 pixels is four times as long as the size of 512 × 512 pixels of SLM3 described above. This is a result of the overall up-conversion to 4 × 4 by performing 2 × 2 up-conversion after 2 × 2 oversampling.

  In this embodiment, the up-converted two-dimensional signal (image) shown in FIG. 7 is processed to extract the X-direction clock and the Y-direction clock included in the image, and based on these extracted clocks, Specify the resampling position. That is, the position of each pixel of the SLM 3 in the image is specified.

For this purpose, first, an X-direction differentiation process and a Y-direction differentiation process are performed on the two-dimensional signal after up-conversion.
FIG. 8 shows an image processing mask used in the differentiation process in this case. FIG. 8A shows an X-direction differentiation mask used for X-direction differentiation, and FIG. 8B shows a Y-direction differentiation mask used for Y-direction differentiation.
As shown in the figure, these masks have a two-dimensional impulse response with the center position as the origin, and are subjected to differential processing in the X direction and Y direction by convolution with the image, respectively. Become.
It is well known that an operator may be called in addition to an image processing mask. Further, these masks are X-direction differentiation and Y-direction differentiation constituting a Sobel operator in image processing, which will be described later, and are often used in differentiation processing.

  Specifically, the size of these masks is a size of a total of 9 pixels of 3 × 3, and the pattern of the X direction differential mask is “−1, 0 in the order of lines (rows) in the X direction from the top. , +1 ”,“ −2, 0, +2 ”,“ −1, 0, +1 ”. The Y-direction differential mask pattern is a pattern in which lines (columns) in the Y direction are set to “−1, 0, +1”, “−2, 0, +2”, “−1, 0, +1” in order from the left side. Become.

  As differentiation processing using these masks, in this example, X-direction differentiation using an X-direction differentiation mask and Y-direction differentiation using a Y-direction differentiation mask are performed independently. That is, the two-dimensional image signal after up-conversion is distributed to two systems, one of which is convolved with an X-direction differential mask and the other is convolved with a Y-direction differential mask. That is, as a result of the differentiation process in this case, two systems of a two-dimensional image signal differentiated in the X direction and a two-dimensional image signal differentiated in the Y direction are obtained.

  As specific contents of the differentiation processing, with respect to the two-dimensional image signal after up-conversion, the center of the mask is aligned with the target pixel, and then the value of the target pixel and the surrounding pixels are calculated. Multiply each value by the corresponding position in the mask. A total of nine values obtained as a result are added up, and the result is set as a differential processing result in the target pixel. This process is performed on each pixel of the two-dimensional image signal after up-conversion.

As an image obtained as a result of such differentiation processing, an image in which the absolute value of the amplitude increases as the change in the amplitude value (that is, the luminance gradient) increases in the image after up-conversion (that is, the original image). Is obtained. In other words, the differential image obtained in this way indicates that the portion having a larger absolute value has a larger luminance gradient.
Such a portion with a large luminance gradient is called an edge. This edge component is an important information source for clock extraction.

Since each mask shown in FIG. 8 includes a smoothing effect, it is necessary to consider the balance with other filter processing. In addition to those listed here, there are a number of masks having a differential effect, which can be appropriately selected and used.
Further, the masks mentioned here are odd number × odd number of 3 × 3 size, and the position (phase) of the pixel of interest is not shifted after convolution. This is preferable because it is not necessary to consider alignment separately.

3-2. Non-linear processing

By the above differentiation processing, a two-dimensional image signal subjected to X-direction differentiation and a two-dimensional image signal subjected to Y-direction differentiation are obtained.
These two-dimensional image signals are further subjected to nonlinear processing. In this case, as nonlinear processing, processing for obtaining an absolute value (absolute value processing) is performed.

9 and 10 show images as timing pulse signals obtained after such absolute value processing. 9 shows an X-direction timing pulse signal obtained by performing absolute value processing on the two-dimensional image signal after X-direction differentiation, and FIG. 10 shows an absolute-value processing on the two-dimensional image signal after Y-direction differentiation. A Y-direction timing pulse signal is shown.
First, as can be seen from these drawings, the X-direction timing pulse signal and the Y-direction timing pulse signal are not binary signals but multi-value signals (more specifically, gray images). Of course, binarization may be used, but in this embodiment, clock extraction is performed from a multilevel signal. The reason is that the sampling rate is relatively low at 4x4, so that the phase information (timing information) is held well by keeping the intensity and waveform of the edge signal as multi-level without binarization. This is to make it happen. In later processing, clock signals are appropriately extracted as fitting them.

Then, as can be seen by comparing FIG. 9 with the previous images of FIG. 6 and FIG. 7, depending on the X-direction differentiation / absolute value processing, white to black and black to white in the X direction in the original image. And the location where the luminance changes is extracted. That is, the edge portion in the X direction is extracted.
Similarly, comparing FIG. 10 with the previous FIGS. 6 and 7, it can be seen that the Y-direction edge portion in the original image is extracted by the Y-direction differential / absolute value processing.

3-3. Two-dimensional Fourier transform processing

For each of the X-direction timing pulse signal obtained by the X-direction differentiation / absolute value processing and the Y-direction timing pulse signal obtained by the Y-direction differentiation / absolute value processing as described above, two-dimensionally Each frequency analysis is performed by performing the Fourier transform processing. From this analysis result, a clock component in the X direction and a clock component in the Y direction can be extracted.

  In the present embodiment, FFT (Fast Fourier Transform) is performed as the Fourier transform. As is well known, FFT is an algorithm for obtaining the same result as DFT (Discrete Fourier Transform) at high speed.

11 and 12 show the analysis result of the X-direction timing pulse signal by the two-dimensional FFT and the analysis result of the Y-direction timing pulse signal by the two-dimensional FFT.
Here, the X-direction timing pulse signal and the Y-direction timing pulse signal each have an image size of 2048 × 2048 pixels. When this is two-dimensionally FFTed, it becomes a complex number array having a size of 2048 × 2048.
First, in the following, the concept of the two-dimensional FFT will be described prior to the description of the analysis result by the two-dimensional FFT.

<Definition>
The definitions of DFT and IDFT (Inverse DFT: Inverse Discrete Fourier Transform), which are the basis of FFT and IFFT (Inverse FFT: Inverse Fast Fourier Transform), are as shown in [Formula 1] and [Formula 2] below.

In the above equation, M is the X direction image size, and N is the Y direction image size. Here, both are 2048 pixels.
F (x, y) represents an image of 2048 × 2048 pixels. Take the x-axis to the right and the y-axis to the bottom,
an integer x = 0, 1,…, 2047,
y is an integer of 0, 1, ..., 2047, and each coordinate has a gray value.
F (fx, fy) is the conversion result of the two-dimensional FFT. This is a complex number.
fx and fy are frequency domain variables, fx is the X direction frequency, fy is the Y direction frequency,
an integer of fx = 0, 1,…, 2047,
It is an integer of fy = 0, 1, ..., 2047.

<Frequency>
Basically, the number of cycles of the sine wave per unit length should be defined. Here, in order to facilitate understanding of the following description, a given image size is assumed to be a unit length. Taking the X direction frequency as an example, the number of one sine wave cycle when the image is advanced in the X direction by the horizontal width (M pixels, that is, 2048 pixels) is the X direction frequency. Similarly, the frequency in the Y direction is the number of cycles when N pixels (2048 pixels) advance in the Y direction.

<Positive and negative frequencies>
fx and fy are non-negative integers as defined above, but their upper half coincides with the negative frequency component.
[Proof: In Equation 1, if fx = −k and fx = M−k (k is an integer) are substituted, x is an integer.
F (−k, fy) = F (M−k, fy)
It turns out that it is equal. ]
Considering the negative frequency, it is easy to understand because the frequency is symmetrical with the origin, and when the bird's-eye view of the power spectrum is viewed later, it is consistent with the frequency being set to 0 at the center. Is considered.
Therefore,
fx = 0, 1,…, 1023, 1024,…, 2046, 2047
The upper half,
fx = 0, 1,…, 1023, -1024,…, -2, -1
As such, it is considered a negative frequency, and the upper half is moved to the front,
fx = -1024,…, -1, 0, +1,…, +1023
It may be understood by converting to an order of frequencies.
The same applies to fy.
Therefore, the frequency analysis result of 2048 × 2048 points can be obtained at any time by performing index conversion of an array suitable for implementation.
fx = -1024,…, -1, 0, +1,…, +1023
fy = -1024,…, -1, 0, +1,…, +1023
It shall be understood as the analysis result of positive and negative frequencies.
The frequency +1024 may be left as +1024 instead of negative -1024, but here it is assumed to be negative.

<Meaning of F (fx, fy)>
As in [Formula 2] above, the image can be decomposed into various frequency components, and conversely, it is expressed as the sum thereof. The frequency component is each term in Σ of [Expression 2], and is expressed by the following [Expression 3].

Here, the exponential portion is a plane wave with X-direction frequency = fx and Y-direction frequency = fy.
F (fx, fy) gives the intensity and phase of the plane wave.

<Complex conjugate component>
The previous [Equation 3] is a complex number, but here, since a grayscale image that is a real value is subjected to frequency analysis, the negative frequency component F (−fx, −fy) of the origin target is always F (fx, fy). It becomes a complex conjugate, and when the sum of the two is taken, the imaginary part disappears and becomes a real number. Therefore, the complex number in [Equation 3] may be considered as a frequency component as it is, and only the positive frequency needs to be noted. The real part may be taken when an individual waveform of the frequency component is actually required.

<Plane wave>
When actually considered as a real number, a plane wave represented by the following [Expression 4] is obtained.

Here, A = | F (fx, fy) |, and θ is an argument of F (fx, fy). A should be doubled if the complex conjugate component of F (fx, fy) is included, but in this embodiment, the absolute value of the amplitude is not a problem and is ignored.

<Cycle>
The wavefront of the plane wave is a straight line, the normal direction of which is the direction of the vector (fx / M, fy / N), and the period L is expressed by the following [Equation 5].

<Frequency analysis>
In this way, frequency analysis is performed using a two-dimensional FFT, and a given grayscale image is decomposed into components that are plane waves of various frequencies (fx, fy) according to [Equation 1] You can know the breakdown. Then, as shown in [Formula 2], if the sum of all components is taken, the original image is restored.

Returning to FIG. 11 and FIG.
FIG. 11 is an overhead view showing the intensity of each frequency component (the square of F (fx, fy)) obtained by two-dimensional FFT of the X direction timing pulse signal shown in FIG. In FIG. 12, the intensity of each frequency component obtained by two-dimensional FFT of the Y direction timing pulse signal shown in FIG. 10 is similarly shown in an overhead view.
As shown in these drawings, in the analysis result of the two-dimensional FFT, the frequency axis is two axes of fx and fy, and the origin is the intersection of both axes. Further, fx and fy have positive and negative frequencies, respectively.
Note that the result of the two-dimensional FFT is 2048 × 2048 points. However, since there is too much information if it is left as it is, it becomes an unclear figure, and in FIGS. 11 and 12, the resolution is reduced to 1/32. . As a matter of course, it is not necessary to reduce the resolution in the original internal processing.
Further, here, for convenience of illustration, only the intensity is shown as a power spectrum, and the phase information of F (fx, fy) as a complex number is not shown, but it is handled as a complex number as internal processing, and the phase information is also missing. Note that it is handled without

In the case of FIG. 11, a large peak stands at a frequency of fx = 512, fy = 0 and a position symmetrical to the origin. The peak portions of fx = 512 and fy = 0 are clock signal components in the X direction. The period of the plane wave corresponding to this frequency is obtained by the above [Equation 5]. Since M = N = 2048, the period L is L = 4. This is derived from the oversampling / up-conversion rate 4 × 4 in this case. Also, the direction of the plane of the wave of this plane wave coincides with the X axis because the Y component is “0” from (512, 0).
This makes sense. In the first place, the image of FIG. 9 which is an input image for the two-dimensional FFT processing includes a large number of edge components in the X direction. The basic interval in the X direction of the X direction edge is 4, and the distance between many arbitrary edges is a multiple of the basic interval 4 even if the distance is short to long. This is because the total magnification by oversampling and up-conversion is 4 times, and the position where the edge is generated becomes the boundary of the data pixels of the SLM 3 (that is, the separation in units of data pixels). Such outstanding features produce peaks at fx = 512, fy = 0 in the analysis of FIG.

Here, if the IFFT of the sum of all the analyzed frequency components is taken, it returns to the original X-direction timing pulse signal of FIG. 9, and if the components are limited, the waveform of that component is obtained.
If only the center component of the peak portion is IFFT, a plane wave as a clock signal corresponding to the waveform of the center component of the X direction edge signal, which is an X direction timing pulse signal, can be obtained. In other words, each edge is sparsely generated in the image, but the center component is a waveform having a period of 4 and the phase is synchronized with the position where the edge is generated. .
Further, if IFFT is performed on the center component of the peak portion and its peripheral frequency components, a plane wave as a clock signal corresponding to the waveform of the main component of the X direction edge signal can be obtained. When IFFT is performed including not only one frequency component giving a peak but also surrounding frequency components, a plane wave close to that is obtained instead of a single plane wave. The peripheral frequency component is a sideband component, and slightly changes the amplitude and phase of a single plane wave, and the change reflects jitter in the image. Therefore, by performing IFFT including this sideband, it is possible to obtain a clock signal that more accurately reflects various fluctuations (enlargement / reduction, rotation, distortion) of the image.

  In the above description, the overall magnification of oversampling / up-conversion is 4 times, the period L is 4, the X-direction clock signal component is fx = 512, fy = 0, and so on. However, this is a design value and actually varies from this value. That is, the original purpose is to generate a clock signal that follows this variation.

Further, the analysis result for the Y-direction timing pulse signal in FIG. 12 is the same as that in FIG. 11 except that the X direction is replaced with the Y direction. That is, the peak portion of the power spectrum is generated with fx = 0 and fy = 512 as the center. This peak component becomes a clock signal component in the Y direction.

3-4. Extraction of clock signal components

<Search range of clock signal component>
When searching for the peak portion from the analysis result by the two-dimensional FFT as described above, a predetermined search range is determined in advance.
Specifically, such a search range is a predetermined range centered at a point of fx = 512, fy = 0 in the analysis result of the two-dimensional FFT in the X direction shown in FIG. 11, and also shown in FIG. In the analysis result of the two-dimensional FFT in the Y direction, a predetermined range centered on a point of fx = 0 and fy = 512 is set.

Here, for confirmation, in hologram recording / reproduction, it is determined how many pixels on the image sensor 6 will receive an image of one pixel of the SLM 3, and the SLM 3 that generates signal light is determined. Since the number of effective pixels is known, it is known in advance how many pixels of the SLM 3 enter from the end of the image sensor 6 in the X direction / Y direction. For this reason, the position where the peak portion appears in the frequency analysis result can be estimated to some extent from the information. Specifically, in the hologram recording / reproduction, the optical system is basically designed so that the range of all effective pixels of the image sensor 6 and the range of all effective pixels of the SLM 3 are matched. Is previously known to contain 512 data pixels in the X direction / Y direction on the image sensor 6. That is, in the analysis result in the X direction, the point of fx = 512, fy = 0 is an ideal peak position, and in the analysis result in the Y direction, fx = 0, fy = 512 is the ideal peak position.
However, in practice, fluctuations such as enlargement / reduction, rotation, and distortion occur in the reproduced image, and accordingly, the peak position appears at a position deviated from the ideal point as a reference. At this time, the distance from the origin of the peak position is the frequency of the clock signal, and the direction of the peak position with respect to the origin coincides with the normal direction of the plane wave as the clock signal. For example, as a typical example, when the reproduced image is enlarged / reduced, the frequency of the clock signal decreases / increases, and the distance from the origin to the peak position decreases / increases. When the reproduced image is rotated, the normal direction of the plane wave as the clock signal is also rotated by the same angle, and the peak position is shifted off-axis.
Note that the results shown in FIGS. 11 and 12 are based on the premise of a clean image with no fluctuations, and therefore there are peaks at ideal positions of fx = 512, fy = 0 and fx = 0, fy = 512, respectively. It is shown that.

  In consideration of these, in this example, as described above, the analysis result of the X-direction timing pulse signal is set to a peak search range based on fx = 512, fy = 0, and the Y-direction timing pulse With regard to the signal analysis results, the peak range is searched for using a predetermined range based on fx = 0 and fy = 512 as a peak search range.

There is a trade-off in setting the search range size. For example, if the search range is too narrow, the range corresponding to fluctuations in the reproduced image tends to be narrowed, but the peak search error can be reduced. On the other hand, if it is too wide, the corresponding range may tend to widen, but the risk of detecting an erroneous peak increases.
In view of these, in this example, a rectangular area of about ± 10% of the reproduction image range centered on each reference point is set as the search range. Since the reference point values are fx = 512 and fy = 512, 10% of them are approximately 50 (512 × 0.1). Therefore, in this case, a rectangular area of 101 (50 + 1 + 50) × 101 (50 + 1 + 50) is defined. It is set as a search range.
The size of the search range is not limited to the above size and can be set arbitrarily. The shape of the search range is not limited to a rectangle, and may be another shape such as a circle.

<Extraction of X clock signal component and Y clock signal component>
As a result of the above search, each peak portion is detected from the analysis result in the X direction and the search result in the Y direction. Subsequently, the clock signal components in the X direction and the Y direction are extracted together with the center component of the detected peak portion and the surrounding components.
In this embodiment, the combination of the center component of the peak portion detected in this way and the surrounding components is referred to as an “X clock signal component” and a “Y clock signal component”, respectively.

13 and 14 show the extraction result of the X clock signal component and the extraction result of the Y clock signal component, respectively.
As shown in these figures, for the peak portion detected by the search, a rectangular region having a size of, for example, 11 × 11 is extracted with the center as a reference. In this case, an area of 2 × 5 + 1 = 11 is set, in which the sideband size is 5, and the plus side and the minus side are combined with one central point as a reference.

Note that the shape of such an extraction region is not rectangular but can be other shapes such as a circle. Further, there is a trade-off with respect to the size of the extraction region, and it may be determined appropriately according to the system design. That is, if the size is small, it becomes difficult to cope with the positional variation in the image, but the effect of not being disturbed by noise is enhanced. Also, if the size is large, it will cope well with positional variations, but conversely it will react to noise and be more likely to be disturbed.

3-5. Phase shift processing

<Jω _x multiplication, jω _y multiplication>
As will be described later, the X clock signal component and the Y clock signal component extracted as described above are processed by IFFT to convert them into real images to obtain an X clock signal and a Y clock signal. However, in each clock signal obtained as a result of IFFT of each extracted clock signal component as it is, the edge timing is represented by the peak portion of the amplitude, so that it is difficult to handle when sampling the edge timing to be performed later End up. For this reason, phase shift processing, specifically differential processing is performed so that the edge timing can be obtained at a zero-cross timing that is easier to handle.

FIG. 15 is a diagram for explaining the phase shift processing, and shows respective waveforms of the sensor output image, the X-direction timing pulse signal, the X-direction clock signal before the X-direction differentiation, and the X-direction clock signal after the differentiation. . Since it is difficult to understand in the two-dimensional state as it is, in FIG. 15, the waveform of each signal is cut and shown as a one-dimensional signal. That is, the waveform of each signal in the figure shows a waveform of a cross section obtained by cutting a two-dimensional image (signal) along a plane perpendicular to the Y axis, the horizontal axis represents the X axis, and the vertical axis represents the gray value (amplitude value). ing.
Although FIG. 15 shows only the X direction, the same applies to the Y direction.

First, as understood from the above description, the X-direction timing pulse signal has a waveform such that the peak is obtained at a portion where the luminance gradient of the sensor output image is high as shown in the figure.
Then, when such an X-direction timing pulse signal is subjected to two-dimensional FFT to extract a clock signal component and IFFT, a cosine wave as shown as an X-clock signal before X-direction differentiation in the figure is reproduced. .

At this time, the ideal sample position is the center of the data pixel and the position of the vertical solid line in the figure. As shown in the figure, the position of the clock signal waveform before differentiation in the X direction is a negative peak position.
It is only necessary that the negative peak position can be detected in the subsequent processing. However, from the viewpoint of ease of detection, it is preferable that the detection at the zero-cross position is possible. Therefore, the cosine wave is differentiated to shift the phase.

Here, such differentiation processing can be performed in the real image region (that is, after IFFT), but here, in consideration of the ease of calculation, processing equivalent to differentiation is performed in the frequency region.
Differentiation in the frequency domain is equivalent to multiplying by an imaginary number jω. Therefore, the clock signal component in the frequency domain obtained in the previous stage is multiplied by jω. Multiply jω in accordance with the frequency of each component in the extracted clock signal component.
The direction of differentiation varies depending on the clock signal component. The X clock signal component is differentiated in the X direction, and the Y clock signal component is differentiated in the Y direction. Accordingly, the X-direction angular frequency jω _x and the Y-direction angular frequency jω _y are multiplied.

By performing such multiplication of jω on the clock signal component extracted in the frequency domain, the phase of the clock signal obtained after IFFT, for example, the waveform of the differentiated X clock signal shown in FIG. Can be shifted so that the zero cross position (positive zero cross position: negative → positive) is the optimum sample position.

3-6. Inverse Fourier transform processing

<X clock signal, Y clock signal>
In the case of the present embodiment, a clock signal based on an actual image is obtained by performing inverse Fourier transform on the clock signal component in the frequency domain obtained by peak search from the analysis result by two-dimensional Fourier transform as described above. It is said.
In this case, since the frequency analysis is performed by FFT, IFFT is performed as the inverse Fourier transform process. As a specific process, the X clock signal component and the Y clock signal component multiplied by jω are IFFT respectively as described above.

FIGS. 16 and 17 show an image obtained by IFFT of the X clock signal component and an image obtained by IFFT of the Y clock signal component, respectively. In this case as well, the signal level is expressed in shades, indicating that the value increases from black to white.
As can be understood from the above description, these images have the wave period, phase, and normal direction as the main components in the X and Y directions for the edge position and intensity included in the sensor output image, respectively. Contains information. In the present embodiment, the images obtained by performing the inverse Fourier transform on the peak components (X clock signal component / Y clock signal component) extracted from the analysis result by the two-dimensional Fourier transform in this way are respectively represented by the X clock. Signal, called Y clock signal.

As described above, since multiplication of jω is performed in the frequency domain, a positive zero cross position (black → white) in these images gives a position to be sampled.
That is, in the X clock signal shown in FIG. 16, the positive zero-cross position represents a break (unit position of each data pixel) in units of data pixels in the X direction. Similarly, also in the Y clock signal shown in FIG. 17, the positive zero cross position represents a delimiter in units of data pixels in the Y direction. In fact, this can be understood by comparing the images in FIGS. 16 and 17 with respect to FIG. 7 (and FIG. 6).

As described above, according to the X clock signal and the Y clock signal, the positive zero-crossing line indicates that the sample position in the data pixel unit in the X direction and the sample position in the data pixel unit in the Y direction in the image, respectively. Can be identified. In other words, the positive zero cross line in the X clock signal becomes a line (X direction periodic line) representing the data pixel period in the X direction in the original reproduced image, and the positive zero cross line in the Y clock signal is the original It becomes a line (Y direction periodic line) representing a data pixel period in the Y direction in the reproduced image.
Therefore, as described later, the positive sampling point of these X clock signal and Y clock signal is extracted, and the intersection point between them is obtained, whereby the resampling position can be specified.

<Resolution in X direction and Y direction when restoring clock signal>
Here, in performing the process of extracting the zero-cross line, the process of changing the resolution in the X direction and the Y direction is performed. That is, prior to searching the images of FIGS. 16 and 17 to obtain the zero cross position, the resolution in the direction of searching for the zero cross position is increased and the resolution in the other direction is decreased. Specifically, with respect to the X clock signal, the resolution in the X direction is increased and the resolution in the Y direction is decreased. For the Y clock signal, the resolution in the Y direction is increased and the resolution in the X direction is decreased.
The purpose of increasing the resolution is to enable easy and accurate processing for searching and determining the zero-cross position. If the resolution is not increased, since the oversample rate is 4 × 4, the basic period of the clock signal is about 4 pixels which is a design standard value on the image data. That is, the waveform of one cycle of the clock signal is expressed by a gray value of about 4 pixels. Extracting a positive zero-cross position from such a signal waveform is not impossible but not easy. Therefore, the resolution is increased several times so that an accurate result can be obtained by an easy process.
On the other hand, the purpose of reducing the resolution in the direction other than the direction in which the zero cross position is searched is to prevent an increase in the amount of calculation accompanying the increase in the resolution.
At this time, if the resolution is not reduced in the other direction, the oversample rate is 4 × 4. Therefore, in the direction, zero cross position information is calculated at a rate of about four per data pixel. . This is excessive and may be a fraction of this. In this direction, it does not actually occur that the sample position of the data pixel is short or fluctuates so much that the zero-cross position has to be determined with a period shorter than the period of the data pixel. This is because the zero-cross line need not be able to express such fluctuations.
For this reason, in this embodiment, when the clock signal component is converted into an image signal by IFFT, the resolution in the direction in which the zero cross position is searched is enlarged, and the resolution is lowered in the other directions.
Specifically, in the case of this example, the resolution in the X direction is expanded four times (2048 × 4 = 8192) for the X clock signal, and the resolution in the Y direction is expanded four times for the Y clock signal.
For the X clock signal, the resolution in the Y direction is reduced to ¼ (2048 ÷ 4 = 512), and for the Y clock signal, the resolution in the X direction is reduced to ¼.
It should be noted that IFFT with different resolutions in the X direction and the Y direction can be realized very easily. More specifically, when the resolution in the X direction is increased for the X clock signal, the number of points in the fx direction is increased to four times in the frequency domain, and the increase is zero. fill in. Further, when the resolution in the Y direction is lowered, the number of points in the fy direction of the frequency domain is set to 1/4 of the low frequency side. Since this result is 8192 × 512 points, an X clock signal can be generated by IFFT.
The Y clock signal may be IFFT after the number of points is adjusted to 512 × 8192 in the same manner.

In this way, by increasing the resolution in the X direction for the X clock signal and increasing the resolution in the Y direction for the Y clock signal, the detection accuracy of the sample position in the X direction and the detection accuracy of the sample position in the Y direction are improved. Can be made.
In addition, for example, if one is reduced by a factor of 4 and the other is reduced to ¼ as in the above specific example, the processing burden is that when performing a normal IFFT without increasing the resolution. It can be suppressed equally.

In this case, as can be understood from the above-described resolution enlargement ratio, the oversample rate for the original image (512 × 512 size) can be 16 times in both the X direction and the Y direction. In other words, the zero cross position can be detected with a resolution 16 times the data pixel unit in the original image.

3-7. Zero cross line extraction

After obtaining an X clock signal and a Y clock signal by IFFT as described above, a positive zero cross line is extracted from them.
FIG. 18 is a diagram for explaining a zero-cross line extraction method.
By the resolution enlargement process, the X clock signal is formed of 8192 sample values in one row in the X direction, and 512 rows in the Y direction are constituted by one row in the X direction. Similarly, in the Y clock signal, one column in the Y direction is formed by 8192 sample values, and one column in the Y direction is configured by 512 in the X direction.
FIG. 18A shows the sample values of each row and each column as a waveform. That is, in the case of an X clock signal, each of the waveforms shown in this figure includes a sample value for one row in the X direction, and there are 512 in the Y direction. In the case of a Y clock signal, each waveform in the figure includes sample values for one column in the Y direction, and there are 512 in the X direction.

  The positive zero-cross line is obtained by extracting a positive zero-cross point for each row in the X direction for the X clock signal, and for each waveform in each column in the Y direction for the Y clock signal, and then obtaining the zero cross obtained in each row and each column. It can be understood as being formed by connecting dots.

FIG. 18B schematically shows a method for extracting a positive zero-cross point.
As shown in the figure, 16 sample values (sampling values) exist within one period of the waveform corresponding to the period of one data pixel by the resolution enlargement process.
The extraction of the positive zero cross point is specifically performed using linear interpolation. In other words, as shown by the dotted circles in the figure, two points that sandwich the polarity change point from negative to positive are found from the waveform of each row or column, and the intersection of the straight line connecting the sampling values of these two points and the zero level Are extracted as positive zero cross points.

Such positive zero-cross line extraction processing is performed for each row for the X clock signal and for each column for the Y clock signal. Then, when the positive zero cross points obtained in each row for the X clock signal are connected in the Y direction, a positive zero cross line for the X clock signal as shown in FIG. 19 is obtained.
Further, when the positive zero cross points obtained in each column for the Y clock signal are connected in the X direction, a positive zero cross line for the Y clock signal as shown in FIG. 20 is obtained.
In FIGS. 19 and 20, each positive zero-cross line is indicated by a solid line, and an image of an X clock signal and an image of a Y clock signal similar to those shown in FIGS. Show.

<Data representation format of clock signal zero-cross line>
By the above processing, the positive zero cross line group of the X clock signal and the positive zero cross line group of the Y clock signal are obtained, and the results are stored in the array variables as follows.
For example, the X clock signal is as follows.
clkx_timing (i, j): The size is 512 × 512 and a real variable.
Meaning: Indicates the positive zero-cross position (real number) of the j-th X clock signal from the left in the row of y = i.
That is, the Y coordinate is an integer coordinate, and the positive zero cross position of the X direction clock is held as a real number that is not an integer. By doing so, it is possible to hold the timing line group in the X direction without any problem in accuracy while adapting to the decrease in the Y direction of the resolution.

The same applies to the positive zero cross line group of the Y clock signal. That is, only x and y are exchanged. Specifically, it is stored in an array variable as follows.
clky_timing (i, j): The size is 512 × 512 and a real variable.
Meaning: Indicates the positive zero-cross position (real number) of the i-th Y clock signal from the top in the column of x = j.

In addition, if the timing line group is held in the above-described expression format, information on what number each zero-cross line is in the whole is also grasped, so as described later, in this way With respect to each resampling position obtained from the stored grid points of each zero-cross line, it is possible to correctly grasp the order relationship before and after in both the X direction and the Y direction. That is, by this, it is possible to obtain the respective pixel amplitude values obtained from the respective resampling positions (that is, the positions of the respective data pixels) while maintaining the order relation before and after that.
This means that the phenomenon called cycle slip, which has occurred in the clock recovery method using the PLL circuit method in the conventional storage apparatus, does not occur.

<About zero cross point extraction processing>
Specifically, the process of storing in the data format as described above while detecting the zero cross point of each clock signal is preferably performed as follows.
First, one zero cross position is found near the position estimated as the center position of the signal light area A2. Based on this, a process of tracing the zero cross position vertically or horizontally is performed.
In this way, if the detection of the zero cross point is expanded from the vicinity of the center to the periphery, a correct zero cross line group in which the order relation of the front, rear, left and right of the zero cross position in the signal light area can be easily obtained. It can be extracted reliably. This can be understood by considering the relationship between the image sensor 6 and the signal light.
For example, as shown in FIG. 22, the shape of the image sensor 6 is a square shape, and the signal light area A2 has a substantially circular shape. Therefore, in the image obtained as the X clock signal, the surrounding portion is signal light. Therefore, it can be expected that the recorded data will reflect the background of the low gray value. Therefore, there is a high possibility that the zero cross point obtained around the image as the X clock signal and the Y clock signal has low reliability.
Therefore, by performing the extraction of the zero-cross points as described above so as to gradually expand from the vicinity of the center of the image, a correct zero-cross line group in which the order relationship of the front, rear, left and right of the zero-cross position in the signal light area matches. Can be easily and reliably extracted.

3-8. Zero-cross line grid point extraction

With the above processing, each zero-cross line group of the X clock signal and the Y clock signal and the position of each zero-cross line in the X direction / Y direction in the image (specifically, what in 512 lines) Information on whether it is the main line).
After that, the position of each data pixel, that is, the resampling position, can be specified by obtaining the intersection (grid point) between each zero cross line group of the X clock signal and each zero cross line group of the Y clock signal. .

  FIG. 21 shows each grid point where both zero cross line groups intersect with a black circle. In FIG. 21, the zero cross line of the X clock signal and the zero cross line of the Y clock signal shown in FIGS.

<Resampling to 1x1>
As described above, the resampling position is specified by each lattice point. After that, if the gray value (amplitude value) of the reproduced image at the resampling position specified from this lattice point on the image after up-conversion shown in FIG. 7 is obtained, the amplitude value in units of data pixels of the SLM 3 Is obtained. That is, 1 × 1 resampling is completed.
Note that the resampling process based on the information on the resampling position specified in this way has already been described as the process of the resampling unit 23 in FIG. 5 described above, and a description thereof is omitted here. Also, the flow of data reproduction after the resampling process has already been described with reference to FIG.

For reference, FIG. 23 shows an image obtained as a result of the resampling process.

4). Internal structure of resampling position specifying unit

Next, referring back to FIG. 5, the internal configuration of the resampling position specifying unit 22 for realizing the resampling position specifying operation as the embodiment described above will be described.
5, the resampling position specifying unit 22 includes an X direction differentiation processing unit 30x, a Y direction differentiation processing unit 30y, an X absolute value processing unit 31x, a Y absolute value processing unit 31y, and an X-FFT as illustrated. Processing unit 32x, Y-FFT processing unit 32y, X clock signal extraction unit 33x, Y clock signal extraction unit 33y, X phase shift processing unit 34x, Y phase shift processing unit 34y, X-IFFT processing unit 35x, Y-IFFT processing A unit 35y, an X zero cross line extraction unit 36x, a Y zero cross line extraction unit 36y, and a zero cross line grid point extraction unit 37 are provided.

  First, the X-direction differentiation processing unit 30x and the Y-direction differentiation processing unit 30y each input an image from the up-conversion unit 21, and execute the X-direction differentiation processing and the Y-direction differentiation processing described above, respectively. That is, the X-direction differentiation processing unit 30x performs the X-direction differentiation processing using the X-direction differentiation mask as shown in FIG. 8A, and the Y-direction differentiation processing unit 30y shown in FIG. 8B. A Y-direction differentiation process using a Y-direction differentiation mask is performed.

The X absolute value processing unit 31x performs processing for converting each value in the image after the X direction differentiation processing by the X direction differentiation processing unit 30x into an absolute value, and supplies the result to the X-FFT processing unit 32x. To do.
The Y absolute value processing unit 31y performs a process of converting each value in the image after the Y direction differentiation processing by the Y direction differentiation processing unit 30y into an absolute value, and outputs the result to the Y-FFT processing unit 32y. Supply.

The X-FFT processing unit 32x and the Y-FFT processing unit 32y are images after absolute value processing (that is, X-direction timing pulse signals, Y, respectively) supplied from the X absolute value processing unit 31x and the Y absolute value processing unit 31y, respectively. A two-dimensional FFT process is performed on the direction timing pulse signal.
As a result, the analysis results as shown in FIGS. 11 and 12 are obtained.

  The X clock signal component extraction unit 33x has a predetermined search range centered on the reference points fx = 512 and fy = 0 as described above for the analysis result obtained by the two-dimensional FFT by the X-FFT processing unit 32x. The peak portion of the power spectrum is searched for within (101 × 101 rectangular region). Then, the central component of the peak portion detected as a result and the peripheral components thereof are extracted as X clock signal components. That is, as described above, an 11 × 11 rectangular region with the center of the detected peak portion as a reference is extracted as an X clock signal component.

  Similarly, the Y-clock signal component extraction unit 33y has an analysis result obtained as a result of the two-dimensional FFT by the Y-FFT processing unit 32y within a predetermined search range centered on reference points fx = 0 and fy = 512 ( The peak portion of the power spectrum is searched in the (101 × 101 rectangular region), and an 11 × 11 rectangular region based on the center of the detected peak portion is extracted as a Y clock signal component.

The X phase shift processing unit 34x multiplies the X clock signal component extracted by the X clock signal component extraction unit 33x by jω. That is, with respect to the X clock signal component, the frequency of each component is set so that the phase of the X clock signal obtained by IFFT of the X clock signal component is shifted as described in FIG. The corresponding X direction angular frequency jω _x is multiplied.

Similarly, the Y phase shift processing unit 34y shifts the phase of the Y clock signal obtained by IFFT of the Y clock signal component extracted by the Y clock signal component extraction unit 33y as described with reference to FIG. As described above, the Y clock signal component is multiplied by the Y-direction angular frequency jω _y corresponding to the frequency of each component.

The X-IFFT processing unit 35x converts the X clock signal component processed by the X phase shift processing unit 34x into an X clock signal as an actual image by IFFT.
Similarly, the Y-IFFT processing unit 35y obtains a Y clock signal by performing IFFT on the Y clock signal component processed by the Y phase shift processing unit 34y.
As can be understood from the above description, the X-IFFT processing unit 35x in this case performs IFFT so that the resolution in the X direction is four times and the resolution in the Y direction is ¼. The Y-IFFT processing unit 35y performs IFFT so that the resolution in the Y direction is quadrupled and the resolution in the X direction is 1/4.

The X zero cross line extraction unit 36x detects the positive zero cross point of each row in the X direction from the X clock signal obtained by the X-IFFT processing unit 35x by the method described in FIG. Are stored in the array variable (clkx_timing (i, j)) as described in (1).
As described above, the zero-cross point is first extracted from the vicinity of the position estimated as the center position of the signal light area A2, and then gradually expanded from there.

  Similarly, the Y zero cross line extraction unit 36y detects the positive zero cross point of each column in the Y direction from the Y clock signal obtained by the Y-IFFT processing unit 35y by the method described in FIG. This is stored by the array variable (clky_timing (i, j)) described above. The extraction of the zero cross point of the Y clock signal is also performed first from the vicinity of the position estimated as the center position of the signal light area A2, and then gradually expanded from there.

The zero cross line lattice point extraction unit 37 extracts the intersection (grid point) of each zero cross line obtained from the extraction result of each zero cross point of the X zero cross line extraction unit 36x and the Y zero cross line extraction unit 36y.
Here, in the state where each zero cross point is stored by the array variable in each zero cross line extraction unit 36 as described above, it remains a set of zero cross points for each row and each column. A set of zero cross points for each row and each column can be treated as a zero cross line. Specifically, information on each zero-cross line is obtained by linearly interpolating a set of zero-cross points stored for each row and each column.
The zero cross line lattice point extraction unit 37 performs such processing to obtain each zero cross line group of the X clock signal and each zero cross line group of the Y clock signal, and then extracts each intersection (each lattice point). To do. From these grid points, the position of the SLM 3 in the data pixel unit in the image is obtained. That is, the resampling position is obtained.

  In addition, by storing each zero cross point by the above array variable, it is possible to grasp the number of each line in the X direction / Y direction as each zero cross line generated as described above from the set of zero cross points. . That is, even with the position of each data pixel specified as a grid point of such a zero-cross line group, it is possible to grasp the number of grid points in the X direction / Y direction in the image.

  The information on the resampling position obtained by the zero-cross line grid point extraction unit 37 is supplied to the resampling unit 23, whereby the amplitude value of each data pixel is obtained as described above. .

  In the above description, the processing for generating the zero cross line group of the X clock signal and the zero cross line group of the Y clock signal by interpolation from the set of zero cross points for each row and each column stored by the array variable is a zero cross line grid. Although described as being performed by the point extraction unit 37, this process may be performed by the X zero cross line extraction unit 36x and the Y zero cross line extraction unit 36y, respectively.

Further, in FIG. 5, for convenience of explanation, it has been described that the image from the up-conversion unit 21 is branched and supplied to the resampling position specifying unit 22 and the resampling unit 23. The image processed by the converting unit 21 may be stored in a memory, and the resampling position specifying unit 22 and the resampling unit 23 may share the image in the memory.

5. Effects of the embodiment

As described above, in the present embodiment, analysis is performed by performing two-dimensional Fourier transform on the two-dimensional image data obtained according to the hologram page data (which is a two-dimensional image) recorded on the hologram recording medium 5. Based on the result of extracting the peak portion of the power spectrum within the first predetermined range and the peak portion of the power spectrum within the second predetermined range from the result, the wave of the clock component in the X direction in the image An X clock signal (clock information) as information representing the period, phase and direction of the image, and a Y clock signal (clock information) as information representing the wave period, phase and direction of the component of the clock in the Y direction in the image Can be obtained respectively. Based on the X clock information and the Y clock signal, the position of the pixel unit in the hologram page data recorded on the hologram recording medium 5 can be specified, and the value of each data pixel is determined based on the position information. Can be acquired.

  At this time, the X clock signal and the Y clock signal acquired on the basis of the analysis result have directions of the clock component in the X direction and the clock component in the Y direction together with the period and phase of the wave as described above. Information is also included. From this, the position of each pixel is specified based on the X clock signal and the Y clock signal as described above, so that the position of each pixel can be specified appropriately in accordance with the rotation of the image. it can. Of course, each pixel position can also be specified appropriately in response to enlargement / reduction of the image. Furthermore, for example, each pixel position can be specified appropriately corresponding to a distortion in which the clock frequency is different between the X direction and the Y direction.

  In addition, according to the resampling technique as the present embodiment as described above, the position of each pixel can be specified (that is, resampling is performed) without performing alignment by subpage sync as in the past. . According to this, it is possible to eliminate the need for a subpage sync for specifying a resampling position, which has been necessary in the past, and to increase the recording capacity of user data accordingly. That is, it is possible to improve the recording density of user data.

  Also, according to the resampling method of the present embodiment, it is possible to specify each resampling position using not only the sync portion in the image but also the entire image data including user data. In other words, conventionally, when a part of an image as a subpage sync is damaged, the position of each pixel is not properly specified in that area, and there is a possibility that the data in that area is completely destroyed. However, according to this embodiment, since the resampling position can be specified using the entire acquired two-dimensional image data, more robust data reading in that sense is possible.

  Also, if the amplitude value of each pixel can be obtained without specifying the position using the subpage sync as described above, the position where each bit value corresponds to the format is determined after bit synchronization. Therefore, it is possible to adopt a normal sync usage method in which frame synchronization is performed by a sync for performing the sync.

In addition, as can be seen from the fact that it is not necessary to insert a subpage sync for specifying the resampling position as described above, according to the present embodiment, restrictions on format formulation are more relaxed than before. can do.
In addition, as described above, since the resampling position is specified using the data of the entire image without distinguishing between the sync and the user data, no limitation is imposed on the recording modulation code.
From these facts, according to the present embodiment, the degree of freedom in format design can be significantly improved as compared with the prior art.

Further, in the present embodiment, when extracting the X clock signal component and the Y clock signal component, the respective images are subjected to differentiation processing to emphasize edges, and then subjected to nonlinear processing (absolute value processing). Thus, the analysis by the two-dimensional FFT is performed.
As a result, the peak level of each clock signal component obtained on the analysis result can be obtained more reliably, thereby making it possible to firmly prevent erroneous detection of the clock signal component.

  Further, in the embodiment, the image after up-conversion is distributed to two systems, and each of the X direction and the Y direction is independent of [X direction differentiation / nonlinear processing / two-dimensional FFT / X clock signal component extraction], [Y-direction differentiation / non-linear processing / two-dimensional FFT / Y clock signal component extraction] is performed, but by doing so, clock signal components are well extracted without interfering with each other. be able to. That is, it is possible to obtain a clock signal with higher accuracy as the X clock signal and the Y clock signal.

  In the embodiment, when the clock signal component is extracted, the clock signal is generated by performing IFFT by combining the center portion of the peak portion of the spectrum and its peripheral components together. By including the band, it is possible to reproduce the clock signal including subtle fluctuations in the actual reproduced image, which may be expressed as jitter. In other words, in the present embodiment in which the resampling position is specified based on such a clock signal, it is possible to specify the resampling position with high accuracy that can follow even subtle fluctuations in the reproduced image.

  Further, in view of the fact that it is possible to follow even a small fluctuation of the reproduced image as described above and that the reproduction image can be enlarged / reduced and rotated as described above, according to the present embodiment, It is not necessary to perform pixel matching in a strict sense that the pixels on the image sensor 6 side and the elements of the SLM 3 are exactly matched. According to this, the degree of freedom in designing the system and the degree of freedom in designing the optical system can be increased, and an increase in cost for improving accuracy can be suppressed.

In the above description, in the present embodiment, it has been described that the resampling position can be specified without using the subpage sync that has been conventionally used. However, as can be understood from the above description, each resampling position is However, at that time, it is still not known which position in the hologram page corresponding to the signal light area A2 those resampling positions are.
Therefore, as described above, alignment by the page sync is performed to specify the position on the format (page sync alignment unit 24). The page sync necessary for such alignment on the format is used. As shown in FIG. 24, only one can be used.

Alternatively, a plurality of page syncs can be inserted as shown in FIG.
When a plurality of page syncs are inserted as shown in FIG. 25, the page sync alignment unit 24 integrates these plurality of page syncs by arranging them even though the locations are distributed. It is handled as a page sync (template), and alignment is performed by template matching by correlation calculation.
Further, when a plurality of page syncs are inserted, they are integrated and function as a page sync as a whole, so there is no problem even if the patterns of individual page syncs are different. In order to improve the correlation characteristics of the page sync as a whole, it can be done intentionally.
Here, when there is one page sync, the information of the entire page cannot be decoded when the sync is damaged. On the other hand, when the page syncs are distributed in a plurality of places as described above, even if some of them are damaged, the page alignment processing result is not affected, and the entire page is not affected. There is a merit that information is not lost.

Here, in the conventional method, the size of the page sync is increased to several times the size of the subpage sync for the convenience of adopting the method of detecting the position of the subpage after first positioning the entire page. The page sync is reliably detected and the entire page is positioned. Specifically, as shown in FIG. 48, the page sync has a size corresponding to one subpage, that is, 9 times the size of the subpage sync. Then, after positioning the entire page, the sub-page sync size is several times larger than the symbol size for the purpose of adopting a method of detecting the position of the sub-page sync and identifying each resampling position from there. Thus, the detection accuracy is increased, and specifically, the size of the subpage sync is four times the size of the symbol as shown in FIG.
As a result, in the conventional method, the size of the page sync is 36 times as large as the symbol.
In the present embodiment, a page sync for confirming the position on the final format is required as described above. In this example, each resampling position is already specified at the time of such alignment. Re-sampling is also performed, and alignment is performed on the re-sampled data, so that it is not necessary to increase the page sync size compared to the symbol size as in the prior art.
As described above, the size of the page sync used in this embodiment can be made smaller than that of the conventional one, and the recording density of user data can be improved in this respect as well.

6). Experimental result

Next, FIGS. 26 to 29 show experimental results for demonstrating the effectiveness of the resampling technique according to this embodiment.

First, prior to the description of the experimental results, the recording format set in conducting the experiment will be described with reference to FIGS.
In conducting the experiment, several improvements were made to the conventional format from the viewpoint of improving the recording density of user data as described above.

First, as a first improvement, the data capacity that can be laid in the circular signal light area A2 is increased by making the minimum unit of data laying in the signal light area A2 smaller than the conventional unit.
Specifically, one symbol as shown in FIG. 30 (in this case, 4 bits × 4 bits = 16 bits, which is the minimum unit of recording coding) is set as the minimum spread unit. That is, the minimum padding unit is reduced to 1/36 from one subpage (= 6 symbols × 6 symbols = 24 bits × 24 bits) which is the conventional minimum padding unit shown in FIG. .

As a second improvement, the data capacity is increased by reducing the ratio of the number of bits occupied by the sink.
First, the size of the sink itself is reduced as shown in FIG. Specifically, the sync size is reduced to 1 symbol with respect to the conventional subpage sync = 2 symbols × 2 symbols.
The sync data pattern may be any pattern as long as it is a combination that does not occur in general recording data. In the experiment, for example, [0 0 0 0,1 1 0 0, 0 0 1 1, 0 0 0 0] was adopted.

Furthermore, an improvement is added to such a sync insertion method.
That is, as shown in FIG. 31B, a two-dimensional arrangement interval (that is, an interval in the vertical and horizontal directions) of each sink is inserted as a predetermined interval i_sper. In FIG. 31B, the periphery of the four syncs when combined with the recording data is shown enlarged.
As shown in FIG. 32, in the experiment, each sync is inserted with a sync interval i_sper = 48 pixels (bits). The sync interval in the conventional format is 6 symbols as shown in FIG. 47, which corresponds to i_sper = 24 pixels (bits). Therefore, by increasing the sync interval, the number of syncs arranged is greatly reduced. It will be.
In FIG. 32, the arrangement position of each sync in the signal light area A2 is indicated by ■. As shown in the figure, each sync is arranged such that one symbol at the center of the page is always the sync position. I went.
In this case, the radius i_rad of the signal light area A2 is set to i_rad = 169 pixels as illustrated.
As a result, the effective data capacity that can be spread on one page is 5555 symbols (bytes), and the recording density is increased compared to the conventional number of effective symbols = 3552 symbols.

  For confirmation, the reason why the radius i_rad of the signal light area A2 can be increased than before is that the minimum spread unit is made smaller than before. In other words, in the past, where the minimum spread unit was set as a subpage unit, the minimum unit that can be increased or decreased for the radius i_rad increases as the minimum spread unit increases, which increases the degree of freedom of the value that can be set as the radius i_rad when defining the recording format. It becomes difficult to ensure. On the other hand, according to the above format in which symbols smaller than the subpage are set as the minimum spread unit, the number of data spreads in the radial direction can be easily adjusted, and as a result, the radius i_rad can be increased. It was made.

Return to the explanation of the experimental results.
FIG. 26 shows the error distribution, SER (symbol error rate), SNR (S / N) in the page when data is reproduced by the conventional reproducing method for the hologram recording medium 5 on which data recording is performed by the above format. Ratio).
FIG. 27 shows error distribution, SER (symbol error rate), and SNR (S / N ratio) in a page when data is reproduced by the reproduction method of the present embodiment.
In these figures, errors in the page are indicated by white squares in the figure, and for the part that is a double square, the inner frame indicates an error in bit units (bit error), The outer frame indicates an error (symbol error) in symbol units.

First, in the case of the conventional reproduction method shown in FIG. 26, it can be confirmed that, for example, in the area in the upper right part of the page, the entire subpage has an error. This indicates that errors have occurred collectively around the sinks that have failed to be detected.
In this case, the SER was 1.27E-001 at 704 symbols / 5555 symbols. The SNR was 2.40.

  On the other hand, in the case of the present embodiment shown in FIG. 27, there is no concentration of errors due to failure of sync detection as in the prior art, and the occurrence positions of errors are distributed and the number thereof is reduced. I understand. In this case, the SER was 7.72E-002 at 429 symbols / 5555 symbols, and the SNR was 2.79.

  From these results, it can be understood that the reproduction signal quality higher than the conventional one can be obtained according to the present embodiment. In other words, in this embodiment, it is possible to suppress a decrease in reproduction signal quality even when the user data has a higher density than before.

Further, FIG. 28 shows margin characteristics with respect to enlargement / reduction of the reproduced image.
In this figure, the horizontal axis represents the image reduction / enlargement ratio, the vertical axis represents the number of errors (the number of symbol errors), and the broken line in the figure shows the result when the conventional reproduction method is adopted, and the solid line represents The result when the reproduction | regeneration system of this Embodiment is employ | adopted is shown.
As shown in the figure, it can be seen that in the conventional case, the number of errors rapidly increases when the enlargement / reduction ratio is about ± 1%. That is, the margin for the conventional enlargement / reduction is only about ± 1%.
On the other hand, in the case of the present embodiment, the number of errors hardly changes until the enlargement / reduction ratio is about ± 9%, and therefore the margin for enlargement / reduction is greatly increased as compared with the conventional case. Recognize.

For reference, FIG. 33 shows an analysis result by a two-dimensional FFT at a magnification of 1.1 times. In FIG. 33 (a), only the vicinity of the X axis is extracted, and the colored portion in the figure indicates the peak search range. In FIG. 33 (b), only the vicinity of the Y axis is extracted, and the colored portion indicates the peak search range. As shown in these figures, the peak component moves in an inversely proportional direction on the axis (that is, a direction in which the frequency decreases with enlargement) depending on the enlargement / reduction of the original image. Will do.
Then, referring to the search range and the peak portion shown in these drawings, it can be seen that the numerical value (about ± 9%) of the previous enlargement / reduction margin is determined by the search range. That is, with respect to the setting of the search range (101 × 101) of the present example described above, the margin is about ± 9%. In other words, if the search range is expanded, further enlargement / The reduction margin can be increased.
This is because, in the characteristic diagram of FIG. 28, in the conventional case, the number of errors increases remarkably from the time when enlargement / reduction occurs, whereas in the case of the present embodiment, the number of errors reaches the margin margin. It is also clear from the fact that there is almost no increase in.

Further, FIG. 29 shows margin characteristics with respect to rotation of a reproduced image.
Also in this figure, the vertical axis represents the number of errors (number of symbol errors), and the horizontal axis represents the rotation angle of the image. The broken line indicates the result when the conventional reproduction method is adopted, and the solid line indicates the result when the reproduction method according to the present embodiment is adopted.
As shown in the figure, it can be seen that the margin for rotation is also greatly improved from the prior art. In the conventional case, the number of errors is noticeably increased as soon as the rotation occurs, and the number of errors rapidly increases when the rotation angle is about 0.5 degrees.
On the other hand, in the case of the present embodiment, it is understood that the number of errors does not substantially change up to about 5 degrees, so that the margin for rotation is increased by about 10 times compared to the conventional case.

FIG. 34 shows an analysis result by two-dimensional FFT at the time of rotation of 5 degrees. Also in FIG. 34, (a) shows only the vicinity of the X-axis, and the peak search range is indicated by colored portions. Further, (b) shows only the vicinity of the Y axis and shows the peak search range in the same colored portion.
As shown in the figure, the peak portion obtained during rotation shifts in the off-axis direction according to the rotation direction of the image without changing the frequency. From such a phenomenon, it can be understood that the direction and amount of off-axis deviation from the reference point of the peak portion in the analysis result represent the direction of rotation and the angle of rotation, respectively.
Also, referring to this figure, it can be seen that the above-described margin value of 5 degrees is limited by the peak search range set in this case. In other words, the margin for rotation can be further increased by expanding the search range.

From the above results, it can be understood that according to the present embodiment, the image is more robust against image deformation / distortion.

7). Modified example

Although one embodiment of the present invention has been described above, the present invention should not be limited to the specific examples described so far.
Hereinafter, each modification will be described.
<First Modification>
In the description so far, in specifying the resampling position, the peak portion is searched from the analysis result of the two-dimensional FFT, and the peak portion is IFFT to obtain a clock signal as a two-dimensional image. The resampling position can be specified without obtaining a clock signal based on a two-dimensional image.

First, in the first modified example, although it is common from the analysis result by the two-dimensional FFT to the peak search, the clock information is directly obtained from the peak portion detected as a result of the search.
As mentioned earlier, the central component of the peak corresponds to a single plane wave that closely approximates the clock signal, and its period, phase, and normal direction are the peak position and value (complex number) in the Fourier domain. Determined from. That is, the distance from the origin of the peak position is the frequency, and its reciprocal is the period. Further, the direction of the peak position with respect to the origin is the normal direction of the plane wave. The value (complex number) at the peak position determines the phase of the plane wave. Therefore, clock information can be obtained from the position and value of the central component of the peak portion obtained as a result of the search.

When the clock information is obtained based on the peak center position in this way, the resampling position can be obtained relatively easily by calculation as follows.

P (m, n) = P0 + m * Lclky + n * Lclkx

However, m and n are integers.
Also,
P (m, n): coordinates of the nth resampling position in the X direction and mth in the y direction,
P0: A reference point of resampling position coordinates, one of the solutions of the optimum sample position obtained from the clock in the X direction and the clock in the Y direction (there are many solutions, but the one near the center of the image is selected and m, n are It is natural to take an integer),
Lclkx: fundamental period vector of clock in X direction (single plane wave),
Lclky: Y-direction clock (single plane wave) basic period vector,
It is.
The basic period vector is a vector having a magnitude equal to the wavelength and a direction matching the wave propagation direction.

  The method for obtaining P0 depends on the timing pulse signal generation method. For example, when the absolute value is obtained by differentiation, the negative peak position is the optimum sample position as shown in FIG. Therefore, it is sufficient to select one near the center of the image.

The point of this first modification is that only the central component of the peak portion obtained as a result of the search for the analysis result is used. In other words, if a single plane wave corresponding to the central component is used as a clock signal, the coordinates of each resampling position are linear as described above based on the period, phase, and normal direction (vector) information. It can be easily obtained by calculation.
According to this, it is not necessary to perform IFFT processing with a relatively large processing load in specifying the resampling position, and the calculation processing load can be greatly reduced.

  Note that the resampling position specifying method according to the first modification example may be performed for the entire page image, but in that case, a single-period plane wave is used as a clock signal. The followability to minute fluctuations will be reduced. Of course, it goes without saying that it is effective for fluctuations such as rotation and enlargement / reduction of the image as a whole.

Here, when it is assumed that a slight fluctuation is dealt with using the technique of the first modification, the image may be divided into a plurality of areas and a series of processes may be performed for each area. By dividing each area in this way, the size of the target range becomes small, and the clock period can be regarded as a single period in each area. That is, by specifying the resampling position by the above linear calculation for each region, it is possible to secure a certain degree of followability to minute fluctuations.
As a specific process, first, a two-dimensional FFT is performed for each region, and a peak search is performed in each region. Then, the re-sampling position is specified by performing the above-described linear calculation based on the peak center component for each region.

  When processing is performed by dividing into regions as described above, the same processing is performed for each region. Therefore, it is possible to perform calculation in parallel by using a plurality of signal processing devices. With such a hardware configuration, the processing time can be greatly shortened.

Note that the above-described region dividing method can also be applied to the case where the IFFT as described above is performed to reproduce a clock signal as a real image. That is, in that case, a two-dimensional FFT is performed for each region, a peak search is performed in each region, and then a peak component is IFFT for each region to obtain a clock signal for each region. Subsequent extraction of the zero-cross line group is performed for each region, but the extraction of the grid points is performed for the entire image using the zero-cross line extracted for each region.
Also in this case, the processing contents can be the same in each region until the zero-cross line group is extracted. Therefore, the processing time can be shortened by adopting a hardware configuration that performs parallel processing on these regions.

  In both cases where the clock signal is handled as a single plane wave and a real image, the hologram page is divided into several areas and the format is defined as an independent set of small pages. It is also possible to reduce the processing time by performing a series of processes in parallel for each small page.

<Second Modification>
In the second modification, the signal light irradiation area (that is, the effective reproduction area) on the image sensor 6 is roughly positioned using the low frequency component of the frequency analysis result. In the embodiment, the timing pulse signal is FFTed to obtain a frequency analysis result. In this process, a low frequency component is also obtained. This is used to obtain a low-resolution image and perform rough positioning based on the low-resolution image.

A second modification will be described with reference to FIG.
FIG. 35A shows a low-resolution image obtained by IFFT of the low frequency component obtained from the frequency analysis result. FIG. 35 (a) shows an example in which a frequency component of 3 or less as a result of the two-dimensional FFT is extracted and converted into a 64 × 64 size image by IFFT.
Since the image size is small, the amount of calculation required here is very small as a whole.

In the second modification, an annular template as shown in FIG. 35B is prepared in advance. This annular template may be understood as a model of a substantially circular shape corresponding to the signal light area A2. That is, the irradiation area of the signal light on the image sensor 6 is specified based on the result of the correlation calculation between the annular template and the image generated as shown in FIG.
As this annular template, the value of the outermost annular part is “−1” (black part in the figure), the value of the annular part adjacent to the inside is “+1” (white part in the figure), and other values. Are all "0" (gray portion in the figure). Since the shape of the signal light area A2 is substantially circular, the alignment is performed only at the edge portion of the grayscale image, aiming to be unaffected by the grayscale unevenness corresponding to the internal recording data. Yes.

  Such a rough positioning process can be performed by using a part of the frequency analysis result, so that the amount of calculation hardly increases. In addition, it is sufficient to perform rough positioning using an image with a low resolution such as the 64 × 64 size as described above. In this respect as well, an increase in calculation amount can be suppressed.

  In realizing the positioning process as the second modified example, first, in the FFT processing unit 32 (any one of 32x and 32y), the low frequency component is extracted together with the search for the peak component. And although illustration is omitted, as a new configuration, the low frequency component extracted by the FFT processing unit 32 is IFFT to obtain a low resolution image as shown in FIG. It is sufficient to add a positioning unit that calculates the position where the maximum value is obtained as position information of the signal light area (effective reproduction area).

If an effective reproduction area is specified, the entire reproduction process can be performed more robustly using the information. An example is shown below.
First, it is conceivable to use it when detecting page sync. That is, based on the position information specified by the positioning unit, the page sync positioning unit 24 sets the search range of the page sync.
If the approximate position of the effective reproduction area is known, the search range for page sync detection can be limited to a narrower range based on the position information. For this reason, it becomes more robust and the amount of calculation can be reduced.

In addition, it may be used when extracting the zero cross point.
According to the above explanation, when extracting the zero-cross point, the extraction starts from the zero-cross point near the reliable center first. However, if the position of the effective reproduction area is specified in advance, it is approximately based on the position information. Therefore, the extraction of the zero cross point may be started from the vicinity of the center position. Specifically, each of the X zero cross line extraction unit 36x and the Y zero cross line extraction unit 36y is configured to start extraction of the zero cross points based on the position information specified by the positioning unit described above.

<Third Modification>
Subsequently, a third modification will be described.
According to the above description, when searching for the peak portion from the analysis result of the two-dimensional FFT, the case where the X clock signal component extracting unit 33x and the Y clock signal component extracting unit 33y independently perform the peak search is illustrated. However, when searching for the peak in the X direction and the Y direction, the peak search in the X direction and the Y direction can be comprehensively performed based on the respective components that are orthogonal to each other.

FIG. 36 is a diagram for describing a third modified example in which such a comprehensive peak search is performed. In this figure, the relationship between the X clock signal component search range and the Y clock signal component search range is shown based on the frequency axis fx and the frequency axis fy.
Here, in the case of this example, since the pixel of the SLM 3 is 512 × 512, the relationship between the clock in the X direction and the clock in the Y direction is almost the same at fx = fy = 512, In addition, the normal vectors of the wave fronts of these plane waves are in a substantially orthogonal relationship, although they are affected by slight fluctuations in the image.
Therefore, a constraint condition that “the clock in the X direction and the clock in the Y direction have the same frequency and the directions of the waves are orthogonal” is imposed, and the search range point in the X direction that satisfies the constraint condition is imposed. For each set of points in the search range in the Y direction, the power spectrum is comprehensively evaluated, and each point giving the maximum evaluation value is defined as a peak in the X direction search range and a peak in the Y direction search range. Get each.

  Specifically, for each set of points in the X-direction search range and points in the Y-direction search range that satisfy the constraint conditions as described above, the sum or product of the power spectrum is calculated as an evaluation value, and the value Is obtained as a peak in the X direction and a peak in the Y direction.

However, such a method is effective when there is almost no distortion of the reproduced image, but there is no guarantee that the relationship between the peak positions is always orthogonal.
Therefore, it is assumed that each peak obtained as described above is a temporary peak, and within this narrower range set based on this temporary peak position again, this time, a more detailed peak search is performed independently for each of the X and Y directions. And the result is obtained as a peak in the final X direction search range and Y direction search range.

  According to such a method, first, a search is performed on the condition of orthogonal relations, so that it can be detected without being confused by the surrounding large false components that may exist as the temporary peak. In addition, by performing a detailed search based on the temporary peak position, peak detection accuracy can be further improved.

In addition, what is necessary is just as follows as a structure for implement | achieving the peak search by the above, for example.
First, for the combination of points in the X direction search range and Y direction search range that satisfy the above constraint conditions, the power spectrum detection order is determined in advance. Each of the X clock signal component extraction unit 33x and the Y clock signal component extraction unit 33y detects the power spectrum of a point in each search range according to the order. Then, either one of these clock signal component extraction units 33 or a newly provided peak determination unit calculates an evaluation value based on the power spectrum of each point obtained as described above, and finally A combination having the maximum evaluation value is acquired as a peak in the X direction search range and a peak in the Y direction search range.

<Fourth Modification>
The fourth modification is an improvement on the differentiation process.
In the description so far, the differentiation processing has branched the image after up-conversion and divided the X-direction differentiation and the Y-direction differentiation to obtain images as independent timing pulse signals. Y-direction differentiation can be simultaneously performed on a common image after up-conversion.

FIG. 37 shows the internal configuration of the data reproducing unit 8 for realizing the fourth modification. In this figure, portions already described in FIG. 5 are given the same reference numerals and description thereof is omitted.
In this case, the X-direction differentiation processing unit 30x, the Y-direction differentiation processing unit 30y, the X absolute value processing unit 31x, the Y absolute value processing unit 31y, the X-FFT processing unit 32x, and the Y-FFT processing unit 32y illustrated in FIG. A Sobel operator processing unit 40 and an FFT processing unit 41 are provided instead.

The Sobel operator processing unit 40 performs Sobel operator processing on the image after up-conversion from the up-conversion unit 21. This Sobel operator processing is equivalent to performing the X-direction differentiation processing and the Y-direction differentiation processing described above simultaneously. Further, according to the Sobel operator process, a process corresponding to the absolute value process is also performed.
The Sobel operator is also described in the following document.
・ "Image recognition theory" Makoto Nagao, Corona, February 15, 1983, first edition

Here, FIG. 38 shows an image as a result of Sobel operator processing.
From this figure, it can be seen that according to Sobel's operator processing, an image in which X-direction edges and Y-direction edges are fused is obtained. This can be well understood by comparing FIG. 38 with the images of the X direction timing pulse signal and the Y direction timing pulse signal shown in FIGS.

  For the image obtained as a result of the Sobel operator processing and corresponding to an image obtained by simultaneously performing X-direction differentiation and Y-direction differentiation (and absolute value processing) on the common image after up-conversion, the FFT processing unit 41 A two-dimensional FFT process is executed.

FIG. 39 shows an analysis result obtained by performing a two-dimensional FFT on the Sobel operator processing result shown in FIG.
When the Sobel operator processing result is two-dimensionally FFT in this manner, it can be seen that there are clearly peaks at positions expected in the design in the vicinity of the X axis and the Y axis on the same Fourier plane. From this, it can be understood that the X-direction clock signal component and the Y-direction clock signal component can be extracted without problems from the analysis result obtained by two-dimensional FFT of the Sobel operator processing result.

  In FIG. 37, the X clock signal component extraction unit 33x and the Y clock signal component extraction unit 33y in this case are determined in advance from the analysis result by the two-dimensional FFT on the Sobel operator processing result obtained by the FFT processing unit 41. The peak search is performed within the search range near the X axis and within the search range near the Y axis. The subsequent processing is the same as that described above.

According to the fourth modification, the two-dimensional FFT processing for clock extraction can be reduced to one time instead of two times, so that there is an advantage that the amount of calculation is greatly reduced.
In the method of the fourth modified example, although the factor that the signals interfere with each other in the X direction and the Y direction increases, the timing pulse signals are basically in a positional relationship orthogonal to each other. There seems to be no problem.

<Fifth Modification>
In the description so far, the timing pulse signal is generated by performing both differential processing and non-linear processing (absolute value processing). However, for the timing pulse signal, only non-linear processing is performed without performing differential processing. It can also be generated.
Here, if differential processing is performed, it is possible to enhance the edge associated with the clock component and promote more reliable clock signal extraction, but the essence of extracting the clock signal component in the frequency domain is to convert the spectrum into an image by nonlinear processing. Accordingly, since the timing pulse signal is generated only by nonlinear processing, clock signal extraction substantially equivalent to that in the embodiment can be performed.

FIG. 40 shows an image obtained by performing absolute value processing on the image after up-conversion as it is.
Further, FIG. 41 shows the analysis result by the two-dimensional FFT for the image after the absolute value processing. As can be seen by comparing FIG. 41 with FIG. 39 above, even when only absolute value processing is performed, the respective reference points near the X axis and Y axis are centered as in the case of performing differentiation processing. As a result, the peak of the power spectrum is obtained.

  FIG. 42 is a diagram for explaining the operation in the case where the timing pulse signal is generated only by the absolute value processing as the fifth modified example, as a diagram for explaining the sensor output image, the timing pulse signal by the absolute value processing, and X Each waveform of the X clock signal before direction integration and the X clock signal after integration is shown. 42, similarly to FIG. 15, the waveform of each signal (each two-dimensional image) is cut into a one-dimensional signal, the horizontal axis represents the X axis, and the vertical axis represents the gray value (amplitude value). It is supposed to be. Although only the X direction is shown in FIG. 42, the same applies to the Y direction.

  First, as absolute value processing, as shown in the figure, a reference level (zero level) for absolute value processing is determined in advance for the sensor output image (actually the image after up-conversion). By performing the absolutization based on this, a timing pulse signal is generated by absolute value processing as shown in the figure. In this case, the reference level may be a local average value or a central value between the local minimum value and the maximum value.

  A waveform of a clock signal obtained by performing two-dimensional FFT and performing peak search, clock signal component extraction, and IFFT on the timing pulse signal by the absolute value processing is indicated by an X clock signal before integration in the X direction in the figure. However, referring to this waveform, unlike the case of FIG. 15, the optimum sample position is a positive peak position. For this reason, as in the case of the previous embodiment, it is necessary to delay the phase when adjusting the optimum sample position to the positive zero-cross position. Therefore, in this case, the phase is shifted by performing integration instead of differentiation. become. Specifically, the clock signal component extracted in the frequency domain is multiplied by an imaginary number −jω so that the optimum sample position becomes a positive zero cross position.

In realizing the operation as the fifth modified example, in the configuration shown in FIG. 37, the image after the up-conversion from the up-conversion unit 21 is absolute for the Sobel operator processing unit 40 instead of the Sobel operator processing unit 40. An absolute value processing unit for performing value processing is provided. The subsequent configuration may be the same as that shown in FIG. 37, but the processing contents of the X phase shift processing unit 34x and the Y phase shift processing unit 34y are different. In other words, the X phase shift processing unit 34x in this case has an X direction angular frequency −jω in the X direction with respect to the X clock signal component obtained in the previous stage so that the positive zero cross position of the waveform after IFFT becomes the optimum sample position. Multiply _x . Further, the Y phase shift processing unit 34y multiplies the Y clock signal component by the Y direction angular frequency −jω _y in the Y direction so that the positive zero cross position of the waveform after IFFT becomes the optimum sample position.

<Sixth Modification>
The sixth modified example treats the image after up-conversion as a timing pulse signal as it is without performing any explicit nonlinear signal processing (differential and absolute value processing, etc.) for generating a timing pulse signal. is there.
Here, according to the above description, as the timing pulse signals in the X direction and the Y direction, the edge in the X direction and the Y direction is emphasized by performing absolute value processing (non-linear processing) on the image after up-conversion. Is done. However, the clock signal component can be extracted by utilizing the nonlinearity hidden in the system without performing such explicit edge enhancement processing.
For example, since the recording / reproduction system has optically and electrically nonlinear input / output characteristics such as the γ characteristic of the two-dimensional light receiving element, the clock signal component can be extracted even if the read signal is subjected to frequency analysis as it is. Can do.

43 and 44 show the analysis results of two-dimensional FFT of the image after up-conversion as it is. FIG. 43 is an enlarged view of the vicinity of the reference point near the X axis (that is, the vicinity of the X clock signal component), and FIG. 44 is an enlarged view of the vicinity of the reference point near the Y axis (the vicinity of the Y clock signal component).
Referring to these figures, when the image after up-conversion is subjected to two-dimensional FFT as it is, it can be confirmed that various frequency components are included and many peaks exist reflecting the original reproduced image. However, if you observe closely, it can be confirmed that each clock signal component is weakly present at the position where it should be.

From these results, it can be seen that the clock signal component can be extracted when the peak search range can be limited to a very narrow range.
For example, the design of the system can be made very strict, and the peak search range can be narrowed when the fluctuation of the image is very small. Therefore, when such a condition is satisfied, the method according to the fifth modification can be suitably applied.

  In realizing the sixth modified example, the Sobel operator processing unit 40 is omitted in the configuration of FIG. 37 so that the image after the up-conversion by the up-conversion unit 21 is directly subjected to the two-dimensional FFT processing. What is necessary is just composition.

<Seventh Modification>
As can be understood from the above description, the two-dimensional FFT is performed when extracting the clock signal component, but the two-dimensional FFT can be realized by repeatedly executing the one-dimensional FFT in the row and column directions. it can. This technique is known as the row-column decomposition method.

FIG. 45 schematically shows a two-dimensional FFT technique based on such a row-column decomposition method. In this way, first, a one-dimensional FFT is performed for each row in the X direction, and then a one-dimensional FFT is performed for each column in the Y direction.
The order of the calculation amount of the one-dimensional FFT is N · log (N). As described above, since the normal row-column decomposition method performs 1-dimensional FFT N times in rows and N times in columns, a total of 2N 1-dimensional FFTs are performed. The order is 2N ² Log (N) in total.

However, the purpose here is to extract the clock component, and therefore it can be considered that the analysis result only needs to be obtained only in the search range of the clock signal component.
Therefore, as shown in FIG. 46, one-dimensional FFT is performed only within a required range for only one of the X direction and the Y direction.
FIG. 46A shows the search range of the X clock signal component and the search range of the Y clock signal component, but the analysis result necessary when extracting the clock signal component is only these portions. Therefore, as shown in FIG. 46B, for example, a one-dimensional FFT is performed on each row in the X direction, and then a one-dimensional FFT is performed only on the columns including the respective search ranges in the Y direction.

  For example, if the width of each search range is 5% of the playback image width, the search range is 10% because the search range is X and Y, and the one-dimensional FFT for each column is reduced to 10%. However, the one-dimensional FFT for each row needs to be performed for all rows, and the total calculation amount is 50% + 50% × 10% = 55%. If ordered, it can be reduced to about ½.

  From the viewpoint of reducing the amount of calculation, a one-dimensional FFT for each row should be able to be reduced in the same way, but since it has already been greatly reduced by the FFT algorithm, it is difficult to significantly reduce the amount of calculation. It is. However, it is only necessary to calculate only the portion related to the target clock signal component near the final stage in the FFT processing, and the calculation amount can be reduced accordingly. If a reduction method for such a row is possible, it can be further applied to the column, and the amount of calculation can be further reduced accordingly.

In the above description, the description is based on the assumption that the one-dimensional FFT is first performed on the rows and then the one-dimensional FFT is performed on the columns. However, when the order is changed, the calculation amount is calculated based on the same idea. Reduction can be achieved.
Further, the technique as the seventh modified example can be applied to a case where a Fourier transform technique other than FFT such as DFT is employed.

For confirmation, the present invention includes a “two-dimensional Fourier transform” including the Fourier transform method according to the seventh modified example and the two-dimensional FFT based on the usual row-column decomposition method. And collectively.
The “Fourier transform” here is not limited to the exemplified FFT, but a two-dimensional vector radix FFT, a method for calculating not only FFT but also DFT as defined, filter bank And any method for performing frequency analysis, such as a method for decomposing into frequency components.

<Other variations>
In the embodiment, the LPF, up-conversion, and differentiation processes have been described as being divided into individual steps. However, this is only for the sake of clarity in explaining the algorithm, and is not limited to this. Is not to be done.
For example, some of the described algorithms can be performed with less computation in the frequency domain. In this regard, a method capable of reducing the amount of calculation may be selected as appropriate according to the actual system.
As a specific example, LPF and differentiation processing can also be performed in the frequency domain. These LPF and differential processing are linear filters and are expressed by convolution operations. It is well known that the convolution operation is equivalent to taking a product for each frequency component in the frequency domain, and the amount of calculation may be small. In addition, since the LPF and differentiation processes can be convolved in advance, the amount of calculation can be reduced accordingly.
In addition, for example, it is well known that up-conversion processing may be performed by extending the size and filling in zero in the frequency domain.
In addition to being able to calculate in advance by combining these things, the amount of calculation can be further reduced by performing calculations in the frequency domain.
However, when the number of taps of the filter is small, the amount of calculation may be smaller when performing convolution in the actual image region.

  In the above description, when extracting the clock signal component on the frequency domain, the X clock signal is searched for a predetermined range near the X axis, and the Y clock signal is searched for a predetermined range near the Y axis. In the situation when the X-axis and Y-axis of the SLM 3 and the image sensor 6 are respectively coincided with each other, this should be done. If it is possible to know the rotation angle and the enlargement ratio in advance as the relationship between the coordinate systems according to the design or operation status, the first predetermined range suitable for extracting the respective clock signal components is determined accordingly. The second predetermined range may be determined, and peak search may be performed using these as the target range to extract each clock signal.

  In the description so far, the description has been made on the premise that up-conversion is performed. For example, when the number of pixels of the image sensor 6 is sufficiently larger than the number of pixels of the SLM 3 and the oversample rate is sufficiently high, etc. There is no need to perform an up-conversion process.

In the previous embodiment, both the positive and negative edges of the differential processing result are used as the timing pulse signal by performing absolute value processing. However, clock extraction is performed from only one edge. be able to. Specifically, the absolute value processing is not performed, and only the positive or negative waveform of the differential result is left, and the other is set to 0.
Although the clock can be extracted by such a method, the merit obtained by using both positive and negative edges as in the case of performing absolute value processing is lost. In other words, if absolute value processing is performed, the clock signal is extracted so as to fit both the positive and negative edges, so that the sample positions are not biased to either direction, and the positive and negative edge positions are extracted. Although the position can be specified with high accuracy assuming that the deviation of the deviation is neutralized, the above method lacks accuracy in this respect.
Further, when absolute value processing is performed, the amount of edge information is relatively doubled, so that the effect of increasing the S / N ratio by increasing the strength of the clock signal component can be obtained.

  For the sake of confirmation, as absolute value processing, processing corresponding to absolute value processing may be performed. For example, processing such as taking the square is included in absolute value processing.

  In the description so far, the case where the sort detection similar to the conventional method is employed in obtaining each bit value from the amplitude value of each pixel finally calculated by resampling (that is, performing data identification) has been exemplified. However, the present invention can also be applied to encoding other than sparse encoding. In this case, a decoding (data identification) method corresponding to the encoding to be used may be employed instead of sort detection. In this case, the decoding uses the amplitude value acquired for each resampling position by the method of the present invention.

In the above description, the case where the present invention is applied only at the time of reproduction of a hologram recording medium has been exemplified. However, in addition to this, for example, a data bit in a required pixel unit such as a two-dimensional barcode reader. The present invention can be suitably applied to any apparatus that reads the value of each pixel in the two-dimensional image on a medium on which data is recorded by the two-dimensional image in which the value is stored.
Usually, a black and white alternating clock pattern is embedded in the two-dimensional barcode to facilitate the processing for obtaining the sample position. However, if the present invention is applied, this can be deleted, and the data capacity Can be increased. In addition, according to the present invention, since resampling can be performed very robustly, it is possible to improve reading performance, such as reducing the error rate as compared with the prior art.
However, for example, in the case of a two-dimensional barcode, unlike in the case of hologram recording / reproduction, it is not determined how many pixels on the image sensor side will receive the image of one pixel in the recorded image. The clock cycle in the Y direction cannot be estimated in advance. Therefore, in such a case, it is difficult to apply a technique that does not perform any explicit nonlinear processing for timing pulse generation as in the sixth modification. That is, in this case, in order to extract the clock frequency component satisfactorily, it is preferable to perform explicit nonlinear processing for generating timing pulses.

  In the above description, the case where the present invention is applied to a recording / playback apparatus capable of both recording and playback has been exemplified. However, the present invention is applied to at least a playback apparatus capable of playback. Can do.

  Further, in the above description, the case corresponding to the reflection type hologram recording medium provided with the reflection film has been exemplified. However, as the present invention, the reproducing apparatus corresponding to the transmission type hologram recording medium not including the reflection film is used. The present invention can also be suitably applied. As described above, a reproducing apparatus that supports a transmissive hologram recording medium can eliminate the need for a beam splitter for guiding diffracted light obtained as reflected light to the image sensor side in accordance with irradiated reference light. . In this case, since the diffracted light obtained in response to the irradiation of the reference light is transmitted through the hologram recording medium itself, an objective lens is further provided on the opposite side of the hologram recording medium when viewed from the laser beam emission point side. The diffracted light as transmitted light may be guided to the image sensor side through the objective lens.

It is the block diagram shown about the internal structure of the hologram reproduction apparatus of embodiment. It is a figure for demonstrating the recording method to a hologram recording medium. It is a figure for demonstrating the reproduction | regeneration method of a hologram recording medium. It is a figure for demonstrating each area of the reference light area, signal light area, and gap area which are prescribed | regulated on a spatial light modulator. It is the block diagram shown about the internal structure of the data reproduction part with which the hologram reproduction apparatus of embodiment is provided. It is the figure which showed the sensor output image. It is the figure which showed the image after up-conversion. It is the figure which showed the image mask for a differentiation process. It is the figure which showed the image as a X direction timing pulse signal. It is the figure which showed the image as a Y direction timing pulse signal. It is the figure which showed the analysis result by 2D FFT about an X direction timing pulse signal. It is the figure which showed the analysis result by two-dimensional FFT about a Y direction timing pulse signal. It is the figure which showed the extraction result of X clock signal component. It is the figure which showed the extraction result of Y clock signal component. It is a figure for demonstrating a phase shift process. It is the figure which showed the image as a X clock signal. It is the figure which showed the image as a Y clock signal. It is a figure for demonstrating the extraction method of the zero cross line of a clock signal. It is the figure which showed the extraction result of the positive zero cross line of X clock signal. It is the figure which showed the extraction result of the positive zero cross line of Y clock signal. It is the figure which showed the extraction result of the grid point of each zero cross line group. It is the figure shown about the positional relationship of an image sensor and signal light. It is the figure which showed the image of the resampling result. It is a figure showing an example of arrangement of a page sync. It is the figure which showed the other example of arrangement | positioning of a page sync. It is the figure which showed each experimental result about the distribution of the error in a page at the time of employ | adopting the conventional reproduction | regeneration system, SER, and SNR. It is the figure which showed each experimental result about the distribution of the error in a page at the time of applying the reproducing | regenerating system of this Embodiment, SER, and SNR. It is the figure which showed the experimental result about the margin characteristic with respect to expansion / reduction of an image. It is the figure which showed the experimental result about the margin characteristic with respect to the rotation of the image. FIG. 4 is a diagram illustrating an example of a minimum unit of data in a signal light area as a diagram for explaining a recording format set when obtaining an experimental result. FIG. 10 is a diagram illustrating an example of a sync pattern and a sync insertion interval as a diagram for explaining a recording format set in obtaining an experimental result. It is the figure which showed the sink arrangement | positioning in a signal light area by the recording format set in obtaining an experimental result. It is the figure which showed the analysis result by two-dimensional FFT at the time of the expansion of an image (1.1 times). It is the figure which showed the analysis result by two-dimensional FFT at the time of image rotation (5 degree | times). It is a figure for demonstrating the 2nd modification. It is a figure for demonstrating the 3rd modification. It is the block diagram which showed the internal structure of the data reproduction part in the case of a 4th modification. It is the figure which showed the operator processing result of Sobel. It is the figure shown about the analysis result obtained by carrying out the two-dimensional FFT of the image after Sobel operator processing. It is a figure which shows the image which performed the absolute value process of the image after up-conversion as it is. It is the figure which showed the analysis result obtained by performing two-dimensional FFT on the image which carried out the absolute value process. It is a figure for demonstrating the phase shift process in the case of a 5th modification. It is the figure which showed the analysis result (X clock signal component vicinity) obtained by carrying out 2D FFT of the image after up-conversion as it is. It is the figure which showed the analysis result (Y clock signal component vicinity) obtained by carrying out 2D FFT of the image after up-conversion as it is. It is the figure which showed typically the method of the two-dimensional FFT by a row-column decomposition method. It is a figure for demonstrating the technique of FFT as a 7th modification. FIG. 6 is a diagram schematically showing a modulation pattern in a spatial light modulator as a diagram for explaining a specific example of a recording data format employed in a conventional recording / reproducing method. As a diagram for explaining a specific example of a recording data format employed in a conventional recording / reproducing method, it is a diagram showing a pattern example of a sync inserted in recording data. It is the figure which showed typically the example of the conventional hologram recording and reproducing method performed based on the conventional recording format. It is a figure for demonstrating the clock extraction method with respect to the conventional one-dimensional image. It is a figure for demonstrating the problem at the time of applying the clock extraction method with respect to the conventional one-dimensional image to a two-dimensional image as it is.

Explanation of symbols

  DESCRIPTION OF SYMBOLS 1 Recording / reproducing apparatus, LD laser diode, 2 Collimator lens, 3 SLM (Spatial light modulator), 4 Beam splitter, 5 Hologram recording medium, 6 Image sensor, OL Objective lens, 7 Data modulation part, 8 Data reproduction part, 20 LPF, 21 up-conversion unit, 22 resampling position specifying unit, 23 resampling unit, 24 page sync alignment unit, 25 symbol extraction unit, 26 data identification unit, 27 sparse code decoding unit, 30x X direction differentiation processing unit, 30y Y-direction differential processing unit, 31x X absolute value processing unit, 31y Y absolute value processing unit, 32x X-FFT processing unit, 32y Y-FFT processing unit, 33x X clock signal component extraction unit, 33y Y clock signal component extraction unit, 34x X phase shift processing unit, 34y Y phase shift processing , 35x X-IFFT processing unit, 35y Y-IFFT processing unit, 36x X zero cross line extraction unit, 36y Y zero cross line extraction unit, 37 zero cross line grid point extraction section, 40 Sobel operator processing unit, 41 FFT processor

Claims (20)

Hologram reproducing apparatus for reproducing hologram recording medium in which data is recorded for each hologram page by interference fringes between reference light and signal light generated by being modulated by a two-dimensional spatial light modulator according to recording data Because
Reference light irradiation means for irradiating the hologram recording medium with the reference light;
Image acquisition means for obtaining two-dimensional image data based on the result of receiving diffracted light corresponding to the hologram page unit data obtained by irradiating the hologram recording medium with the reference light;
Analysis means for performing a two-dimensional Fourier transform process based on the two-dimensional image data and performing a frequency analysis on a plane wave element included in the two-dimensional image data;
A search is made for the peak portion of the power spectrum within the first predetermined range and the second predetermined range with respect to the analysis result by the analyzing means, and based on the peak portion detected within the first predetermined range. First direction clock information indicating the period, phase, and normal direction of the specified plane wave, and the period, phase, and normal direction of the plane wave specified based on the peak portion detected within the second predetermined range. Clock information acquisition means for acquiring second direction clock information representing
Pixel position specifying means for specifying a position of the spatial light modulator in the pixel unit in the two-dimensional image data based on the first direction clock information and the second direction clock information;
Amplitude value acquisition means for acquiring the amplitude value of each pixel based on the position information specified by the pixel position specification means;
A hologram reproducing apparatus comprising: Differential processing means for performing differential processing on the two-dimensional image data obtained by the image acquisition means;
A non-linear processing means for performing non-linear processing on the two-dimensional image data subjected to differential processing by the differential processing means,
The analysis means is
Two-dimensional Fourier transform processing is performed on the two-dimensional image data after the nonlinear processing by the nonlinear processing means.
The hologram reproducing apparatus according to claim 1, wherein: Non-linear processing means for performing non-linear processing on the two-dimensional image data obtained by the image acquisition means,
The analysis means is
Two-dimensional Fourier transform processing is performed on the two-dimensional image data after the nonlinear processing by the nonlinear processing means.
The hologram reproducing apparatus according to claim 1, wherein: The nonlinear processing means includes
As the non-linear process, an absolute value of the two-dimensional image data is taken, or a process of taking a square value is executed.
4. The hologram reproducing apparatus according to claim 2, wherein the hologram reproducing apparatus is characterized in that: The clock information acquisition means
As the first direction clock information and the second direction clock information, a first direction two-dimensional image obtained by performing an inverse Fourier transform on a component based on a peak portion detected within the first predetermined range and converting the component into a real image. Obtaining a clock image and a second direction two-dimensional clock image obtained by performing inverse Fourier transform on the component based on the peak portion detected within the second predetermined range and converting the component into a real image,
The pixel position specifying means includes:
The first direction periodic line represented by the first direction two-dimensional clock image and the second direction periodic line represented by the second direction two-dimensional clock image were extracted, and each lattice point where each periodic line intersected was obtained. In the above, the position in the pixel unit is specified based on the position of each grid point.
The hologram reproducing apparatus according to claim 1, wherein: The clock information acquisition means
As the first direction clock information and the second direction clock information, an inverse Fourier transform is applied to the component including the central component of the peak portion detected in the first predetermined range and its neighboring components. A real image obtained by performing inverse Fourier transform on a component including the first direction two-dimensional clock image converted into an image and the central component of the peak portion detected in the second predetermined range and its neighboring components. Each of the second direction two-dimensional clock image converted into
The pixel position specifying means includes:
The first direction periodic line represented by the first direction two-dimensional clock image and the second direction periodic line represented by the second direction two-dimensional clock image were extracted, and each lattice point where each periodic line intersected was obtained. In the above, the position in the pixel unit is specified based on the position of each grid point.
The hologram reproducing apparatus according to claim 1, wherein: First direction differentiation processing means for performing first direction differentiation processing on the two-dimensional image data obtained by the image acquisition means;
First direction nonlinear processing means for performing nonlinear processing on the data after differential processing by the first direction differential processing means;
Second direction differentiation processing means for performing second direction differentiation processing on the two-dimensional image data obtained by the image acquisition means;
Second direction nonlinear processing means for performing nonlinear processing on the data after differentiation processing by the second direction differentiation processing means,
The analysis means is
Two-dimensional Fourier transform processing is performed individually on the two-dimensional image data after processing by the first direction nonlinear processing means and the two-dimensional image data after processing by the second direction nonlinear processing means,
The clock information acquisition means
The first direction clock information is obtained based on the analysis result of the two-dimensional image data processed by the first direction nonlinear processing means by the analysis means, and processed by the second direction nonlinear processing means by the analysis means. Obtaining the second direction clock information based on the analysis result of the two-dimensional image data;
The hologram reproducing apparatus according to claim 1, wherein: The amplitude value acquisition means includes
Performing an interpolation process using values around the position of each pixel specified by the pixel position specifying means to obtain the amplitude value of each pixel by calculation;
The hologram reproducing apparatus according to claim 1, wherein: The amplitude value acquisition means includes
An interpolation process based on any of the nearest neighbor method, bilinear interpolation method, cubic interpolation method, and bicubic spline method is executed as the interpolation process.
The hologram reproducing apparatus according to claim 8, wherein: In the hologram page unit data, a predetermined data pattern predetermined as a page sync is inserted, and
further,
Based on the result of searching the page sync position in the two-dimensional image data based on the amplitude value of each pixel acquired by the amplitude value acquisition means and the predetermined data pattern as the page sync, Alignment means for specifying the position of each pixel in the format;
Reproduction processing means for performing data reproduction based on the value of each pixel acquired by the amplitude value acquisition means and the position information on the format of each pixel specified by the alignment means,
The hologram reproducing apparatus according to claim 1, wherein: Multiple page syncs are inserted,
The alignment means is
The position of each pixel on the format is specified based on a result of searching the plurality of page syncs as an integrated page sync.
The hologram reproducing apparatus according to claim 10. The clock information acquisition means
When searching for the peak portion based on the analysis result by the analysis means, the frequency is the same and the normal direction is orthogonal as a combination of points within the first predetermined range and the second predetermined range. The power spectrum is detected for each combination of points that are related to each other, and based on the detected power spectrum of each point, an evaluation value that evaluates the overall power spectrum of each point is calculated, and this evaluation is performed. Comprehensively searching for a peak portion within the first predetermined range and a peak portion within the second predetermined range based on a calculation result of the value;
The hologram reproducing apparatus according to claim 1, wherein: The clock information acquisition means
Based on the calculation result of the evaluation value, a comprehensive search of the peak portion in the first predetermined range and the peak portion in the second predetermined range is performed, and the first detected as a result is detected. Search for the peak portion again within the respective ranges set with reference to the positions of the peak portion within the predetermined range and the peak portion within the second predetermined range,
The hologram reproducing apparatus according to claim 12, wherein: Integrated differential processing means for simultaneously performing differential processing in both the first direction and the second direction on the common two-dimensional image data obtained by the image acquisition means;
A non-linear processing unit that performs non-linear processing on the common two-dimensional image data that has been subjected to differential processing in both the first direction and the second direction by the integrated differential processing unit;
The analysis means is
Two-dimensional Fourier transform processing is performed on the common two-dimensional image data subjected to the differential processing in both the first direction and the second direction after processing by the nonlinear processing means.
The hologram reproducing apparatus according to claim 1, wherein: The image acquisition means obtains the two-dimensional image data such that the X direction / Y direction of the spatial light modulator matches the X direction / Y direction of the two-dimensional image data,
The clock information acquisition means
The X-direction clock frequency and the Y-direction clock frequency estimated based on the number of pixels in the X-direction and Y-direction of the spatial light modulator and the sampling rates in the X-direction and Y-direction in the image acquisition means are used as references. In the first predetermined range and the second predetermined range that are respectively set as a search for peak portions for obtaining clock information, respectively.
The hologram reproducing apparatus according to claim 1, wherein: The clock information acquisition means
Obtaining the first direction clock information and the second direction clock information based only on the central component of the peak portion detected in each of the first predetermined range and the second predetermined range,
The pixel position specifying means includes:
The linear calculation based on the information on the period, phase and normal direction of the single plane wave indicated by the first direction clock information, and the information on the period, phase and normal direction of the single plane wave indicated by the second direction clock information. Go to identify the location in pixels,
The hologram reproducing apparatus according to claim 1, wherein: The analysis means is
Performing the two-dimensional Fourier transform individually for each partial region of the two-dimensional image data obtained by the image acquisition means;
The clock information acquisition means
Obtaining the first direction clock information and the second direction clock information for each partial region based on the analysis result for each partial region of the two-dimensional image data obtained by the analysis means;
The pixel position specifying means includes:
Based on the first direction clock information and the second direction clock information for each partial area obtained by the clock information acquisition means, a process for specifying the position of each pixel for each partial area is performed.
The hologram reproducing apparatus according to claim 1, wherein the hologram reproducing apparatus is characterized in that: Hologram reproduction for reproducing a hologram recording medium in which data is recorded for each hologram page by interference fringes between reference light and signal light generated by modulation by a two-dimensional spatial light modulator according to recording data A method,
A reference light irradiation procedure for irradiating the hologram recording medium with the reference light;
An image acquisition procedure for obtaining two-dimensional image data based on the result of receiving diffracted light corresponding to the hologram page unit data obtained by irradiating the hologram recording medium with the reference light,
An analysis procedure for performing a two-dimensional Fourier transform process based on the two-dimensional image data and performing a frequency analysis on a plane wave element included in the two-dimensional image data;
A search is made for the peak portion of the power spectrum within the first predetermined range and the second predetermined range with respect to the analysis result by the analysis procedure, and based on the peak portion detected within the first predetermined range. First direction clock information indicating the period, phase, and normal direction of the specified plane wave, and the period, phase, and normal direction of the plane wave specified based on the peak portion detected within the second predetermined range. Clock information acquisition procedure for acquiring second direction clock information representing
Based on the first direction clock information and the second direction clock information, a pixel position specifying procedure for specifying the position of the spatial light modulator in the pixel unit in the two-dimensional image data;
An amplitude value acquisition procedure for acquiring the amplitude value of each pixel based on the position information specified by the pixel position specification procedure;
A hologram reproducing method comprising: A reading device for reading a value of each pixel in the two-dimensional image with respect to a medium on which data is recorded by a two-dimensional image in which a data bit value is stored in a required pixel unit,
Light irradiation means for irradiating the image on the medium with light;
Image acquisition means for obtaining two-dimensional image data corresponding to the two-dimensional image on the medium based on the result of detecting light obtained in response to light irradiation on the medium;
Nonlinear processing means for applying nonlinear processing to the two-dimensional image data obtained by the image acquisition means;
Analysis means for performing two-dimensional Fourier transform processing on the two-dimensional image data processed by the nonlinear processing means, and performing frequency analysis on plane wave elements included in the two-dimensional image data;
A search is made for the peak portion of the power spectrum in each of the first predetermined range and the second predetermined range with respect to the analysis result by the analyzing means, and the peak portion detected in the first predetermined range is detected. First direction clock information indicating the period, phase, and normal direction of the plane wave specified based on the above, and the period, phase, and normal direction of the plane wave specified based on the peak portion detected within the second predetermined range. Clock information acquisition means for acquiring second direction clock information representing
Based on the first direction clock information and the second direction clock information, pixel position specifying means for specifying a position in units of pixels in the two-dimensional image recorded on the medium in the two-dimensional image data;
Amplitude value acquisition means for acquiring the amplitude value of each pixel based on the position information specified by the pixel position specification means;
A reading device comprising: A reading method for reading a value of each pixel in the two-dimensional image with respect to a medium on which data is recorded by a two-dimensional image in which data bit values are stored in a required pixel unit,
A light irradiation procedure for performing light irradiation on the image on the medium;
An image acquisition procedure for obtaining two-dimensional image data corresponding to the two-dimensional image on the medium based on a result of detecting light obtained in response to light irradiation on the medium;
A non-linear processing procedure for applying non-linear processing to the two-dimensional image data obtained by the image acquisition procedure;
An analysis procedure for performing a two-dimensional Fourier transform on the two-dimensional image data processed by the nonlinear processing procedure, and performing a frequency analysis on a plane wave element included in the two-dimensional image data;
A search is made for the peak portion of the power spectrum within the first predetermined range and the second predetermined range with respect to the analysis result obtained by the analysis procedure, and the peak portion detected within the first predetermined range is detected. First direction clock information indicating the period, phase, and normal direction of the plane wave specified based on the above, and the period, phase, and normal direction of the plane wave specified based on the peak portion detected within the second predetermined range. Clock information acquisition procedure for acquiring second direction clock information representing
Based on the first direction clock information and the second direction clock information, a pixel position specifying procedure for specifying a position in units of pixels in the two-dimensional image recorded on the medium in the two-dimensional image data;
An amplitude value acquisition procedure for acquiring the amplitude value of each pixel based on the position information specified by the pixel position specification procedure;
A reading method comprising: