JP2009118144A - Device for embedding marker for image displacement detection, and image displacement detector - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide a displacement detector capable of detecting displacement with durability even for a still image and a moving image to which lossy compression is performed. <P>SOLUTION: A block division means 11 divides the luminance signal of an input image for each macro block (MB). A marker embedding means 12 selects an MB at a marker embedding position from among MBs and embeds markers in MxN pieces of selected MBs. In this case, the marker embedding positions in the MBs divided into M sets and belonging to the same set are considered as the same, and displacement is provided to the marker embedding positions in the MBs belonging to different sets. A table 13 stores marker embedding position information 13-1 and displacement information 13-2. When the displacement of the image occurs, the displacement of the image is detected using characteristics that energy of a DCT factor after lossless compression is distributed to be low. <P>COPYRIGHT: (C)2009,JPO&INPIT

Description

  The present invention relates to an image position shift detection marker embedding device and an image position shift detection device for detecting a position shift of a still image or a moving image, and more particularly to a still image or a moving image subjected to irreversible compression. The present invention relates to an image position shift detection marker embedding device and an image position shift detection device that can detect a position shift.

  When a still image or a moving image is edited or attacked, the position of the reproduced image may deviate from the position of the original image. The image positional deviation detection technique is a technique for detecting how much positional deviation has occurred in a still image or a moving image in which positional deviation may occur as described above.

  Image position shift detection technology is used to correct the position of an image so that the image is reproduced as intended by the producer (image producer), or to detect a digital watermark.

  Patent Document 1 discloses highly reliable information by extracting secondary information after correcting a geometric transformation applied to an image or signal in which secondary information such as a digital watermark is embedded. An image processing apparatus for performing extraction is described.

  Japanese Patent Application Laid-Open No. H10-228561 describes that a digital watermark for moving images is detected after correctly recovering the synchronization in the spatial direction before detecting the digital watermark.

  Cited Document 3 describes a digital watermark embedding device that can detect embedded watermark information when there is an attack such as image clipping, scaling, and rotation.

Patent Document 4 describes a misregistration detection device that detects misregistration between two images.
JP 2002-354221 A JP 2006-295688 A JP 2002-325233 A JP 2004-145466 A

  However, the image processing apparatus of Patent Document 1 embeds a registration signal together with secondary information and corrects the geometric transformation amount using this registration signal, and the tolerance for lossy compression is taken into consideration. Absent.

  The devices described in the cited documents 2 and 3 detect the positional deviation of the image using the watermark information itself, and even this does not consider the resistance to lossy compression.

  In order to detect the amount of misalignment between two image data, the misregistration detection device of Patent Document 4 obtains a phase at at least one frequency from the image data, and based on this phase difference, the position between the images. The amount of deviation is obtained, and even this does not take into account the resistance to lossy compression.

  An object of the present invention is to provide a position shift detection device that can detect a position shift with tolerance even for still images and moving images that have been subjected to lossy compression.

  In order to solve the above-mentioned problem, an image position deviation detection marker embedding device according to the present invention includes a first unit that divides an input image into blocks having a preset size, and a block that is divided by the first unit. Image position misalignment detection marker embedding device comprising: second means for determining a plurality of blocks in which marker information is embedded from within; and third means for embedding marker information in the plurality of blocks determined by said second means The third means embeds marker information giving a predetermined amount of positional deviation in each of the plurality of blocks determined by the second means, and a block in which marker information is embedded and the block The first feature is that a table indicating the relationship with the positional shift given to the marker information embedded in the block is held.

  Further, the image position shift detection marker embedding device of the present invention has a second feature in that the third means embeds marker information by adding or subtracting base signals of single or plural DCT coefficients.

  In the image position shift detection marker embedding device of the present invention, the third means selects a combination of two sub-blocks in each of the plurality of blocks determined by the second means, and A third feature is that marker information is embedded in either of the sub-blocks.

  In the image position shift detection marker embedding device according to the present invention, when the input image is a moving image, the third means creates a plurality of types of tables, and stores information in the plurality of types of tables for each frame. There is a fourth feature in that marker information is embedded using any one of the above information.

  In the image position shift detection marker embedding device according to the present invention, the third means performs a process of embedding marker information using any one of the information in the plurality of types of tables for each frame. There is a fifth feature in that it is applied in the order of return.

  An image position deviation detection apparatus according to the present invention is an image position deviation detection apparatus that receives an image in which marker information is embedded by the image position deviation detection marker embedding apparatus, and detects the position deviation of the image. And a fourth means for specifying the position of the embedded marker information from the relationship between the block in which the marker information is embedded and the positional deviation given to the marker information embedded in the block, There is a first feature in that a fifth means for calculating a positional deviation of an image based on position information specified by the means is provided.

  Further, in the image position deviation detection device of the present invention, the fifth means detects the mutual relationship with the marker information from the block in which the marker is embedded by the third means at the position specified by the fourth means. The second feature is that it has an extracting means for extracting a correlation value.

  In the image position deviation detection device according to the present invention, the fifth means causes the marker information to be shifted by one line at a time within a range of allowable values of position deviation around the position specified by the fourth means. And the absolute value of the correlation value for each shift amount (x, y) is accumulated, and the shift amount (x, y) at which the accumulation reaches the maximum value is detected as a positional deviation of the image. There is a third feature.

  In the image position deviation detecting device of the present invention, the fifth means uses the position deviation given when the marker information is embedded as an index, and the absolute value of the cross-correlation value with the marker information for each block having the same index. A fourth feature is that a value is accumulated, and a position shift corresponding to an index at which the accumulation has a maximum value is detected as a position shift of the image.

  An image position deviation detection device according to the present invention receives an image in which marker information is embedded by the image position deviation detection marker embedding device having the second feature described above, and the extraction means A fifth feature is that a DCT coefficient is extracted as a cross-correlation value.

  According to the present invention, in the image position deviation detection device having the fifth feature, the fifth means is 1 within a range of allowable values of position deviation centered on the position specified by the fourth means. The DCT coefficient is acquired while shifting one pixel at a time, the absolute value of the DCT coefficient for each shift amount (x, y) is accumulated, and the shift amount (x, y) at which the accumulation reaches the maximum value is used as the image position shift. There is a sixth feature in the point of detection.

  Further, according to the present invention, in the image positional deviation detecting device having the fifth feature, the fifth means uses the positional deviation given when embedding the marker information as an index, and the DCT is performed for each block having the same index. A seventh feature is that the absolute values of the coefficients are accumulated, and the positional deviation corresponding to the index at which the accumulation has the maximum value is detected as the positional deviation of the image.

  The image position deviation detection device of the present invention receives an image in which marker information is embedded by the image position deviation detection marker embedding device having the third feature as input, and the fifth means is configured to embed marker information. Using the given positional deviation as an index, find the match rate between the magnitude of the cross-correlation value with the marker information for the two sub-blocks in the block where the marker information is embedded and the magnitude relation of the correlation value when the marker is embedded The eighth feature is that a cumulative value of the matching rate is obtained for each block having the same index, and a positional shift corresponding to an index having the maximum cumulative value is detected as a positional shift of an image.

  The image position deviation detection device of the present invention receives an image in which marker information is embedded by the image position deviation detection marker embedding device having the third feature as input, and the fifth means is configured to embed marker information. Using the given positional deviation as an index, for each block having the same index, the absolute value difference of the cross-correlation values with the marker information for the two sub-blocks in the block in which the marker information is embedded is accumulated. A ninth feature is that a positional shift corresponding to an index having a maximum value is detected as a positional shift of an image.

  Furthermore, the image position deviation detection device of the present invention has a tenth feature in that the fifth means uses an average value instead of the accumulation.

  In the present invention, marker information for detecting a positional deviation is embedded in an image, and when the positional deviation occurs in the image due to editing or an attack, the mutual information with the marker information after irreversible compression (encoding including quantization) is performed. Since the energy of the related value (DCT coefficient) is dispersed and lowered, it is possible to reliably detect the positional deviation of the image with tolerance even for still images and moving images subjected to lossy compression.

  In the present invention, it is possible to detect a positional shift with high durability even for an image processed by an irreversible compression method in which quantization such as MPEG or JPEG is performed for each predetermined block. A position shift is given to the marker information and it is embedded in the block.

  When a position shift occurs in an image due to editing or attack, the correlation value with the marker information after irreversible compression (encoding including quantization) becomes low because the energy is dispersed according to the position shift. The positional deviation of the image can be detected by extracting.

  The present invention will be described below with reference to the drawings. First, a marker embedding device for detecting an image position shift according to the present invention will be described. In the following, it is assumed that an image in which a marker is embedded is irreversibly compressed by DCT according to MPEG, JPEG, or the like. Therefore, the marker is embedded as a DCT coefficient after lossy compression, and the DCT coefficient becomes a cross-correlation value with the marker. In addition, marker information for detecting image position deviation is simply referred to as a marker.

  FIG. 1 is a block diagram showing a first embodiment of a marker embedding device according to the present invention. The marker embedding device of the first embodiment embeds a marker in a single sub-block. The positional deviation of the image can be detected by utilizing the property that the energy of the DCT coefficient is dispersed and lowered when the positional deviation of the image occurs.

  The marker embedding device according to the first embodiment includes a block dividing unit 11 that divides an input image into macroblocks, a marker embedding unit 12 that embeds a marker in a selected macroblock, marker embedding position information 13-1, and deviation information 13 A table 13 for storing -2 and a shift giving means 14;

  The input image is input to the block dividing means 11. The block dividing means 11 divides the luminance signal of the input image every 16 (pixels) × 16 (lines) corresponding to the macroblock (MB). The input image may be a still image or a moving image.

  The marker embedding unit 12 first selects the MB at the marker embedding position from among the MBs divided by the block dividing unit 11. The table 13 stores marker embedding position information 13-1, and the marker embedding position information 13-1 includes the MB number of the marker embedding position. The marker embedding unit 12 selects the MB of the marker embedding position according to the marker embedding position information 13-1. Here, a total of M × N MBs are selected (M, N> 0). However, it is preferable to select adjacent MBs so that image quality deterioration due to marker embedding is not noticeable.

  Next, the marker embedding unit 12 embeds a marker in the selected M × N MBs. The marker is embedded in one of the sub-blocks SB (8 (pixels) × 8 (line)) constituting the MB. Also, MxN MBs are divided into M sets of N (set 0, set 1,...), The marker embedding positions in MBs belonging to the same set are the same, and MBs belonging to different sets The marker embedding positions at are different from each other. The table 13 stores misalignment information 13-2, and the misalignment information 13-2 includes positional misalignments to be given to MB SBs belonging to each set. The deviation giving means 14 gives a deviation to the marker embedding position when the marker embedding means 12 embeds the marker according to the deviation information 13-2.

  For example, the marker is embedded as the positional deviation (parallel movement amount) of the SB in which the marker is embedded as (0, 0) for the MB belonging to the set 0, (1, 1) for the MB belonging to the set 1,. The marker can be embedded by various methods, for example, by adding or subtracting a base signal of a specific DCT coefficient to the luminance signal of the SB at the position translated in the MB as described above. . In a specific example, the marker is a base signal of DCT coefficient (1, 1), and this is added to the luminance signal of SB. When an image in which this marker is embedded is irreversibly compressed (DCT), the DCT coefficient (1, 1) becomes a value having a cross-correlation with the marker. Adding or subtracting the base signal of the DCT coefficient (1,1) is equivalent to adding or subtracting a constant to the DCT coefficient (1,1). Since the marker is embedded as a watermark, the original image is hardly disturbed.

When M = 8 and N = 10, each set (set 0, set 1,..., Set 7) and the corresponding parallel movement amount are set as follows, for example.
Set 0: (0,0), set 1: (1,1), set 1: (2,2), set 3: (3,3), set 4: (4,4),
Set 5: (-3, -3), Set 6: (-2, -2), Set 7: (-1, -1)

  FIG. 2 shows an example of marker embedding positions for each set (set 0, set 1,...), The grid indicates the MB boundary, the shaded portion indicates the MB of the marker embedding position, and the bold line embeds the marker. SB is shown. Here, set 0: (0,0), set 1: (1,1), set 1: (2,2), set 3: (3,3), set 4: (4,4) are illustrated ing.

  As described above, a still image can be composed of the frame in which the marker is embedded. In the case of a moving image, the above marker embedding is performed over several frames. However, it is preferable to change the MB of the marker embedding position for each frame so that the embedded marker is not conspicuous. Also, if the MB of the marker embedding position is made different for each frame in a predetermined plurality of frames and the same processing is repeated with this as a cycle, the processing load can be reduced. This is because multiple types of tables 13 are created and marker information is embedded using one of the tables for each frame, and one table among the multiple types of tables is rotated for each frame. This can be realized by embedding marker information in the order of return.

  FIG. 3 is a diagram illustrating an example of information stored in the table 13. In the figure, the MB number corresponds to the marker embedding position information 13-1, and indicates the MB of the marker embedding position. The amount of deviation corresponds to the deviation information 13-2, and indicates the positional deviation given to each MB SB at the marker embedding position and the set to which it belongs. A group corresponds to a frame when the processing target is a moving image.

  Here, in the first frame (group 1), MBs with MB numbers 10, 26,... Belonging to the set 0, 2,. 0,0), (2,2), ..., and in the next frame (group 2), the MB of MB number 11, ... belonging to the set 1, ... It is shown that the positional shifts given to (1, 1),.

  In the marker embedding device of the first embodiment, only a single SB of the MB at the marker embedding position is used. However, two SBs in the MB at the marker embedding position are used and both SBs are used. It is also possible to detect a positional deviation of the image. This will be described below as the marker embedding device of the second embodiment.

  In the marker embedding device of the second embodiment, a marker is embedded in one of two focused SBs in the MB at the marker embedding position. This marker embedding is also performed by shifting the embedding position and adding or subtracting a base signal of a specific DCT coefficient as in the first embodiment. When a positional shift occurs in an image with a marker embedded, the total positional difference between the DCT coefficients of both SBs hardly changes, or the positional shift of the image takes advantage of the fact that the distribution of the size comparison between both SBs is uniform. Can be detected. Since the block configuration is the same as that of the first embodiment, the illustration is omitted.

  In the marker embedding of the second embodiment, the same method as that of the first embodiment is adopted, but the marker is embedded in one of any two predetermined SBs in the MB at the marker embedding position. At that time, a marker is embedded in the SB that can embed a larger value. In this way, the difference in absolute value of the DCT coefficient between the SB with the marker embedded and the SB without the marker embedded is made large.

  Next, an image position deviation detection apparatus according to the present invention will be described. FIG. 4 is a block diagram showing a first embodiment of an image position deviation detection device according to the present invention.

  The image position deviation detection apparatus according to the first embodiment uses the property that the energy of the DCT coefficient is dispersed and lowered due to the position deviation of the image, and the image in which the marker is embedded by the marker embedding apparatus in FIG. And

  The image position deviation detecting device of the first embodiment includes a marker position specifying means 41, a deviation calculating means 42, and a table 43 storing marker embedding position information 43-1 and deviation information 43-2. The contents of the table 43 are the same as those of the table 13 of FIG. 1, and therefore the MB at the position where the marker is embedded for each group is also known on the image position deviation detection device side.

  The marker position specifying means 41 uses an image in which a marker is embedded by the marker embedding apparatus of FIG. 1 as an input image, divides the input image into MBs, and stores marker embedded position information 43-1 stored in the table 43. Based on, specify the MB of the marker embedding position.

  The deviation calculation means 42 extracts the DCT coefficients of the SB used for embedding in the MB at the marker embedding position and the SBs in the vicinity thereof, and calculates the position deviation of the image based on this.

  Specifically, the DCT coefficient for an 8 (pixel) x 8 (line) block while shifting by 1 line per pixel within the allowable range of positional deviation around the SB used for embedding in the MB at the marker embedding position Are accumulated, and the absolute values of the DCT coefficients at MxN points obtained for each shift amount (x, y) are accumulated. This is performed over a plurality of frames. Then, the shift amount (x, y) having the maximum accumulated absolute value is detected as an image position shift. This is an SB search process that maximizes the accumulation of absolute values of DCT coefficients.

  5 and 6 are diagrams conceptually showing the processing in the marker position specifying means 41 and the deviation calculating means 42. FIG. 5 shows the case where the image positional deviation is (0, 0), and FIG. This is the case when the deviation is (-2, -2).

  In FIG. 5, the marker position specifying means 41 uses an image in which the marker is embedded as an input image, divides the input image into MBs (FIG. 5A), and also stores the marker embedding position stored in the table 43. The MB at the marker embedding position is specified based on the information 43-1 ((b) in the figure).

  Next, the deviation calculating means 42 is one pixel within the allowable range of deviation around the SB used for embedding in the MB at the marker embedding position (within the search range of FIG. The DCT coefficients for the 8 (pixel) × 8 (line) block (SB) are acquired while shifting one line at a time, and the absolute values of the DCT coefficients at MxN locations acquired for each shift amount (x, y) are accumulated. This is performed over a plurality of frames. Then, the shift amount (x, y) having the maximum accumulated absolute value is detected as an image position shift. In this case, since the accumulation of absolute values of the DCT coefficients at the MxN locations when the shift amount is (0, 0) is the maximum value, the deviation calculating means 42 calculates the positional deviation of the image as (0, 0).

  In FIG. 6, as in FIG. 5, the marker position specifying means 41 divides the input image into MBs (FIG. 6 (a)) and specifies the MB at the marker embedding position (FIG. 5 (b)). Then, the deviation calculating means 42 acquires the DCT coefficient for SB within the search range of FIG. 7C, and accumulates the absolute values of the DCT coefficients at MxN locations acquired for each shift amount (x, y). In this case, since the accumulation of the absolute values of the DCT coefficients at the MxN locations when the shift amount is (−2, −2) is the maximum value, the displacement calculating means 42 sets the positional displacement of the image to (−2, −2). ) And calculate.

  The shift calculation means 42 can use a search processing technique generally employed in motion vector search for moving picture coding. That is, first, roughly, for example, the DCT coefficient for SB is obtained every other pixel, and then the search process is performed while shifting the pixel area by one line around the pixel area where the accumulated absolute value is the maximum value. It is possible to adopt a technique of executing and finally obtaining a shift amount with the maximum accumulated value of DCT coefficients being accumulated. As a result, the number of search processes can be reduced to speed up the shift calculation process. This can be similarly applied to the following embodiments.

  In addition, since the process of acquiring the DCT coefficient is equivalent to the process of acquiring the cross-correlation value with the base signal of the DCT coefficient, if processing such as scaling is performed on the image, an 8x8 By obtaining a scaled base signal of the DCT coefficient in the same manner and obtaining a cross-correlation value therewith, it is possible to detect the positional deviation of the image. This can be similarly applied to the following embodiments.

  When the processing target is a moving image and the marker embedding position is different for each frame and this is periodically performed in a period of several frames, it is necessary to know which group each frame belongs to. It is possible to insert some information for this purpose, but each frame is processed by processing all groups for the first few frames and looking at the distribution of the sum of absolute values for the shift amount (x, y). You can identify which group you belong to. When processing as a correct group, the distribution varies, but when processing as an incorrect group, the distribution becomes uniform. Since the subsequent frames are periodic marker embeddings, it is known to which group these belong. This can be similarly applied to the following embodiments.

  Next, a second embodiment of the image position deviation detection device according to the present invention will be described. The image positional deviation detection apparatus of the second embodiment also uses the property that the energy of the DCT coefficient is dispersed and lowered due to the positional deviation of the image, and the image in which the marker is embedded by the marker embedding apparatus in FIG. And Since the block configuration is the same as that of the first embodiment, the illustration is omitted.

  The image position deviation detection apparatus according to the second embodiment does not perform a search process for calculating the deviation, acquires a DCT coefficient within the predetermined SB of the MB at the embedding position, and calculates an image position deviation based on the DCT coefficient. This is different from the image position deviation detection device of the first embodiment.

  Specifically, the DCT coefficient for the upper left SB of the MB at the marker embedding position is acquired, and the absolute value of the DCT coefficient for the SB to which the same positional deviation is given at the time of embedding is accumulated. Whether or not the SBs have the same positional deviation at the time of embedding can be known from the table storing the marker embedding position information and the deviation information. Based on this table, the positional deviation given at the time of embedding is used as an index, and the absolute values of the DCT coefficients for the SBs having the same index, that is, the same positional deviation given at the time of embedding the marker are accumulated. Then, the positional deviation corresponding to the index having the maximum accumulated absolute value is set as the positional deviation of the image. That is, when the cumulative value of the absolute value for a certain index is larger than the other values, it is determined that the positional shift corresponding to the index that maximizes the cumulative value of the absolute value is the calculated image positional shift.

  FIGS. 7 and 8 are diagrams conceptually showing the shift calculation processing in the second embodiment. FIG. 7 shows that when the image position shift is (0,0), FIG. 2, -2).

  In FIG. 7, first, an image in which a marker is embedded is set as an input image, the input image is divided into MBs (FIG. 7A), and marker embedding is performed based on marker embedding position information stored in the table. The MB of the position is specified ((b) in the figure). Next, the DCT coefficient for the upper left SB of the MB at the marker embedding position is acquired, and the positional deviation given to the SB at the time of marker embedding is used as an index, and the absolute value of the DCT coefficient acquired for each index is accumulated (FIG. c)). This is performed over a plurality of frames. Then, the position shift corresponding to the index having the maximum absolute value accumulation is detected as the image position shift. In this case, since the accumulation of the absolute value of the DCT coefficient for the SB to which the positional deviation of (0,0) is given becomes the maximum value, the positional deviation of the image is detected as (0,0).

  In FIG. 8, as in FIG. 7, the input image is divided for each MB (FIG. 8A), and the MB at the marker embedding position is specified (FIG. 7B). The DCT coefficient for the upper left SB is acquired, and the positional deviation given to the SB at the time of embedding the marker is used as an index, and the absolute value of the DCT coefficient acquired for each index is accumulated ((c) in the figure). In this case, since the accumulation of the absolute value of the DCT coefficient for the SB to which the positional deviation is given by (2, 2) is the maximum value, the positional deviation of the image is calculated as (−2, −2).

  Next, a third embodiment of the image position deviation detection device according to the present invention will be described. The image position deviation detection apparatus of the third embodiment uses the relationship between the difference or distribution of the DCT coefficient energy and the image position deviation for the two SBs of interest at the marker embedding position MB. An image in which a marker is embedded by the marker embedding device of the embodiment is a processing target. Since the block configuration is the same as that of the first embodiment, the illustration is omitted.

  In the shift calculation, the image in which the marker is embedded by the marker embedding device of the second embodiment is processed, the image is divided into MBs, and the marker embedding position is based on the marker embedding position information stored in the table. Based on the deviation information, the positional deviation of the image is calculated based on the difference between the DCT coefficients of the two SBs used for embedding the MB at the marker embedding position.

  Specifically, after obtaining the DCT coefficient of the SB used for embedding in the MB at the marker embedding position, the image position shift is calculated by one of the following two methods (A) and (B). From the marker embedding position information and deviation information in the table, the position deviation given to each MB of the mark embedding position and the SB used for mark embedding can be known.

  (A) This position shift is used as an index. Find the magnitude relationship between the DCT coefficients obtained for two SBs in the MB. Then, the coincidence ratio between the magnitude relation and the magnitude relation when the mark is embedded is obtained, and the cumulative value of the coincidence ratio for each index is obtained. A positional shift corresponding to the index having the maximum cumulative value is detected as a positional shift of the image. That is, when the cumulative value for a certain index is far greater than the other, it is determined that the positional shift corresponding to the index having the maximum cumulative value is the positional shift of the image.

  (B) This positional deviation is used as an index. The absolute value difference of the DCT coefficients acquired for the two SBs in the MB is obtained, and the accumulated absolute value difference for each index is obtained. At this time, the position shift corresponding to the index that maximizes the cumulative value of the absolute value differences is detected as the image position shift. That is, when the cumulative value for a certain index is far greater than the other, it is determined that the positional shift corresponding to the index having the maximum cumulative value is the positional shift of the image.

  In the above embodiment, as the number of sets (the type of positional deviation given when embedding a marker) is increased, the positional deviation of the image can be detected with higher accuracy. However, the processing load increases accordingly.

  Even if image misalignment occurs, if it is a translation in the horizontal (X-axis) or vertical (Y-axis) direction, the variance of the DCT coefficient is relatively small, and it remains relatively large even after quantization. By utilizing the property, it is possible to detect the positional deviation of the image from a small number of sets with a light processing load. For example, the set is a positional shift (0,0), (1,1), ..., (7,7), and first, as described above, find the location where the sum of absolute values of DCT coefficients is maximized, Next, a search process may be performed near the location.

1 is a block diagram illustrating a first embodiment of a marker embedding device according to the present invention. It is a figure which shows the example of the marker embedding position with respect to each set. It is a figure which shows the example of the information which a table stores. 1 is a block diagram showing a first embodiment of an image position deviation detection device according to the present invention. FIG. FIG. 5 is a diagram conceptually showing marker position specification and deviation calculation processing (in the case where an image positional deviation is (0, 0)). FIG. 5 is a diagram conceptually showing marker position specification and deviation calculation processing (in a case where an image positional deviation is (−2, −2)). It is a figure which shows notionally the shift calculation process in 2nd Embodiment of the image position shift detection apparatus which concerns on this invention (when the position shift of an image is (0,0)). FIG. 10 is a diagram conceptually illustrating a shift calculation process in the second embodiment of the image position shift detection device according to the present invention (when the image position shift is (−2, −2)).

Explanation of symbols

11 ... Block dividing means, 12 ... Marker embedding means, 13-1, 43-1 ... Marker embedding position information, 13-2, 43-2 ... Deviation information, 13, 43 Table, 14 ... Deviation providing means, 41 ... Marker position specifying means, 42 ... Deviation calculating means

Claims (15)

First means for dividing the input image into blocks of a predetermined size;
Second means for determining a plurality of blocks in which marker information is embedded from among the blocks delimited by the first means;
An image position deviation detection marker embedding device comprising third means for embedding marker information in a plurality of blocks determined by the second means,
The third means embeds marker information giving a predetermined amount of positional deviation in each of the plurality of blocks determined by the second means, and embeds the block in which the marker information is embedded and the block. A marker embedding device for detecting an image position deviation, characterized by holding a table indicating a relationship with a position deviation given to marker information.   2. The marker embedding device for image position deviation detection according to claim 1, wherein the third means embeds marker information by adding or subtracting base signals of single or plural DCT coefficients.   The third means selects two sub-blocks in each of the plurality of blocks determined by the second means, and embeds marker information in one of the two sub-blocks. 3. A marker embedding device for detecting an image position shift according to 1 or 2.   When the input image is a moving image, the third means creates a plurality of types of the table, and embeds marker information using any one of the information in the plurality of types of tables for each frame. The image position shift detection marker embedding device according to any one of claims 1 to 3.   5. The method according to claim 4, wherein the third means applies processing for embedding marker information using any one of the information in the plurality of types of tables for each frame in the order of return. The marker embedding device for image position deviation detection as described. An image position shift detection device that receives an image in which marker information is embedded by the image position shift detection marker embedding device according to claim 1 and detects the position shift of the image,
A fourth means for identifying the position of the embedded marker information from the relationship between the block embedded with the marker information and the positional shift given to the marker information embedded in the block, which is held in the table;
An image position deviation detecting device comprising: fifth means for calculating an image position deviation based on the position information specified by the fourth means.   The fifth means has extraction means for extracting a cross-correlation value with marker information from a block in which a marker is embedded by the third means at a position specified by the fourth means. The image position shift detection device according to claim 6.   The fifth means obtains a correlation value with the marker information while shifting one pixel at a time within a range of allowable values of positional deviation around the position specified by the fourth means, and a shift amount ( The absolute value of the correlation value for each x, y) is accumulated, and a shift amount (x, y) at which the accumulation takes a maximum value is detected as an image positional deviation. Image position deviation detection device.   The fifth means uses the positional deviation given when the marker information is embedded as an index, accumulates the absolute value of the cross-correlation value with the marker information for each block having the same index, and the accumulation takes the maximum value. The image position deviation detection device according to claim 7, wherein a position deviation corresponding to the index is detected as an image position deviation. An image in which marker information is embedded in the image position deviation detection marker embedding device according to claim 2 is input,
The image position deviation detection device according to claim 7, wherein the extraction unit extracts a DCT coefficient as a cross-correlation value with the marker information.   The fifth means obtains a DCT coefficient while shifting the pixel one line at a time within the allowable range of positional deviation around the position specified by the fourth means for each shift amount (x, y). 11. The image position deviation detection device according to claim 10, wherein absolute values of the DCT coefficients are accumulated, and a shift amount (x, y) at which the accumulation reaches a maximum value is detected as an image position deviation.   The fifth means uses the position shift given when the marker information is embedded as an index, accumulates the absolute value of the DCT coefficient for each block having the same index, and the position corresponding to the index where the accumulation has the maximum value. The image position deviation detection device according to claim 10, wherein the deviation is detected as an image position deviation. An image in which marker information is embedded in the image position deviation detection marker embedding device according to claim 3 is used as an input,
The fifth means uses the positional shift given at the time of embedding the marker information as an index, and the magnitude relationship between the cross-correlation values with the marker information for the two sub-blocks in the block in which the marker information is embedded, and the mutual relationship at the time of marker embedding. The matching rate of the relationship value with the magnitude relationship is obtained, the cumulative value of the matching rate is obtained for each block having the same index, and the positional shift corresponding to the index having the maximum cumulative value is detected as the positional shift of the image. The image position deviation detection device according to claim 7, wherein An image in which marker information is embedded in the image position deviation detection marker embedding device according to claim 3 is used as an input,
The fifth means uses a positional shift given when marker information is embedded as an index, and for each block having the same index, a cross-correlation between the marker information of two sub-blocks in the block in which the marker information is embedded The image position deviation detection device according to claim 7, wherein an absolute value difference between the values is accumulated, and a position deviation corresponding to an index at which the accumulation reaches a maximum value is detected as an image position deviation.   15. The image position deviation detection device according to claim 8, wherein the fifth means uses an average value instead of the accumulation.