JPH11261961A - Moving image processor - Google Patents

Abstract

PROBLEM TO BE SOLVED: To allow the processor to conduct processing of a moving image such as analysis, retrieval, collation without referring to an original image or an image of the same size as that of the original image. SOLUTION: A signature calculation section 102 calculates a signature of a digital moving image in the unit of one field or frame based on prescribed rules. A moving image analysis section 104 uses this signature to conduct moving image analysis processing such as detection of 3:2 pull-down pattern, detection of a still image part, and retrieval and collation of a moving image without referring to the original image. Thus, processing such as analysis, retrieval, and collation of a moving image is conducted by using a storage section with a small capacity, a low speed transmission section and a low speed processing section.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a moving picture processing apparatus for processing digital moving pictures, and more particularly to a moving picture processing apparatus for analyzing, retrieving and collating digital moving pictures.

[0002]

2. Description of the Related Art In recent years, digitization of video has been rapidly progressing with the progress of digital video processing technology and the fusion of various media. Since the video signal has a larger amount of information than the text information and the audio signal, the MPEG2 video standard (I
In many cases, the amount of information is reduced or transmitted by a moving picture coding technique (compression technique) represented by SO / IEC13818-2) and then transmitted or stored.

In the MPEG2 video standard, one screen is divided into a plurality of blocks of 8 pixels × 8 pixels, and each block is subjected to DCT (discrete cosine transform) to obtain a DC.
It is based on quantizing the T coefficient and performing variable length coding. Also, by using the motion compensation inter-picture prediction together, the coding efficiency is enhanced by utilizing the correlation of the images in the time direction.

A moving image compression technique such as MPEG2 is a useful technique for reducing the amount of information and facilitating video handling. However, even if it is compressed data, the information amount is large, especially HDTV (high definition TV) and UD
In the case of a TV (ultra high definition TV), it cannot be easily decoded by software. In moving image processing, for example, moving image analysis, search, and collation, it is not only necessary to refer to or decode compressed data, but also to refer to an original image or an image of the same size as the original image (for example, a decoded image). Processing is often necessary.

[0005] For example, when detecting the correspondence with the original 24 frames from the interlaced signal of 60 fields per second converted from a movie of 24 frames per second (detection of 3: 2 pull-down pattern) or included in a moving image This corresponds to the case where a series of still image portions to be detected is detected.

[0006] Even with the current standard TV, if the original image itself or an image of the same size as the original image is to be referred to or processed,
It is necessary to support a data rate of about 200 Mbps. Furthermore, the data rate of HDTV is about 1G.
bps, and UDTV has to cope with a data rate several times higher than that. Therefore, the required processing speed becomes enormous, and there is a problem that a high-speed transmission path and dedicated processing hardware must be prepared.

[0007]

As described above, in moving image processing, for example, analysis, retrieval, and collation of moving images, it is not only necessary to refer to or decode compressed data, but also to use an original image or an original image. Often, it is necessary to refer to and process images of the same size, which increases the required processing speed and requires the provision of high-speed transmission paths and dedicated processing hardware. there were. Especially in HDTV and UDTV,
This problem was remarkable.

SUMMARY OF THE INVENTION The present invention has been made to solve the above problems, and performs a moving image processing such as analysis, retrieval, and collation of a moving image without referring to or processing an original image or an image of the same size as the original image. It is an object of the present invention to provide a moving image processing device capable of performing the above.

[0009]

In order to solve the above-mentioned problems, a moving image processing apparatus according to the present invention comprises a signature calculating means for calculating a signature based on a predetermined rule for each predetermined unit of a digital moving image; Analyzing means for analyzing the digital moving image based on the image representative information including at least the signature without referring to the digital moving image.

Thus, information representing an image can be generated with an extremely small amount of information irrespective of the size of the screen, and a digital moving image can be analyzed based only on the image representative information.

According to the present invention, there is provided a moving image processing apparatus comprising:
A signature calculating unit for calculating a signature based on a predetermined rule for each predetermined unit of the digital video, and a transmission unit for transmitting image representative information including at least the signature separately from the digital video. Features.

Thus, it is possible to transmit information representing an image with an extremely small amount of information regardless of the screen size without transmitting the original image, that is, the digital moving image itself.

According to the present invention, there is further provided a receiving means for receiving image representative information including at least a signature calculated based on a predetermined rule for each predetermined unit of a digital moving image, and a method for receiving image representative information received by the receiving means. On the basis of,
Analyzing means for analyzing the digital moving image without referring to the digital moving image.

Thus, it is possible to analyze the digital moving image based on only the image representative information having a very small information amount regardless of the screen size, without referring to or processing the original image, that is, the digital moving image itself.

The digital moving picture analyzing means receives image representative information including at least a signature calculated based on a predetermined rule for each field of the digital moving picture, and refers to the digital moving picture without referring to the digital moving picture. It is preferable to include a detecting unit that detects a 3: 2 pull-down pattern of the digital moving image based on a matching state of one frame interval of the signature.

Thus, the 3: 2 pull-down pattern can be reliably detected based only on the coincidence of the one-frame interval of the signature calculated on a field basis.

In another preferred embodiment of the digital moving picture analyzing means, image representative information including at least a signature calculated based on a predetermined rule for each field or frame of the digital moving picture is input, and 1 of the signature without referring to a moving image
A detecting means for detecting whether or not the digital moving image is completely still based on the matching state of the frame intervals.

Thus, the still image portion in the moving image can be reliably detected based only on the coincidence of the one-frame interval of the signature calculated on a field or frame basis.

Further, another preferable example of the digital moving picture analyzing means is that the first moving picture is calculated based on a predetermined rule for each field or frame of the digital moving picture.
Image representative information including at least the signature of the first moving image and the first moving image without referring to the digital moving image.
Search means for searching for the digital moving image based on the coincidence state of the signature of the second and the second signature set in advance.

Thus, by simply setting the second signature, it is possible to reliably search for an image based only on the coincidence between the second signature and the received first signature.

According to the present invention, there is provided a signature calculating means for calculating a signature based on a predetermined rule for each predetermined unit of a digital moving image, and a compressing means for compressing and encoding the digital moving image to generate compressed moving image data. Transmitting means for transmitting image representative information including at least the signature together with the compressed moving image data.

Thus, the compressed data and the image representative information of an extremely small amount of information can be transmitted regardless of the size of the screen without transmitting the original image, that is, the digital moving image itself. And various analysis processes using the above can be efficiently performed.

According to the present invention, there is provided a moving image processing apparatus comprising:
Image representative information including at least a first signature calculated based on a predetermined rule for each predetermined unit of the first digital video, and compressed video data generated by compression-encoding the first digital video Receiving means, a decoding means for decoding the compressed moving image data to obtain a decoded moving image, and calculating a second signature based on the predetermined rule for each of the predetermined units of the second digital moving image Signature calculating means for detecting whether or not the second digital moving image is the same as the first digital moving image based on a coincidence state between the first signature and the second signature. And a detecting means.

Thus, the second information is obtained based on only the image representative information having an extremely small information amount regardless of the screen size.
Can be compared with the first digital moving image which is the original image of the compressed moving image data.

According to the present invention, there is provided a moving image processing apparatus comprising:
Compression means for compressing and encoding a digital moving image to generate compressed moving image data; signature calculating means for calculating a signature based on a predetermined rule for each predetermined unit of a locally decoded moving image of the compressed moving image data; Transmitting means for transmitting the decoded image representative information including at least the signature together with the compressed moving image data.

Thus, without transmitting the original image, that is, the digital moving image itself or the decoded moving image itself,
It is possible to transmit compressed data and local decoded image representative information having an extremely small information amount regardless of the size of the screen.

According to the present invention, there is provided a moving image processing apparatus comprising:
A decoded image representative including at least a first signature calculated based on a predetermined rule for each predetermined unit of compressed moving image data generated by compression-encoding a digital moving image and a locally decoded moving image of the compressed moving image data. Receiving means for receiving information; decoding means for decoding the compressed moving image data to obtain a decoded moving image; and calculating a second signature based on the predetermined rule for each of the predetermined units of the decoded moving image. Signature calculation means; and detection means for detecting whether or not the decoded moving image is normal based on the coincidence state of the first signature and the second signature.

Thus, based on only the very small amount of decoded image representative information regardless of the screen size,
It is possible to confirm whether or not the decoding of the compressed moving image data has been correctly performed.

According to the present invention, there is provided a moving image processing apparatus comprising:
Receiving means for receiving image representative information including at least a first signature calculated based on a predetermined rule for each predetermined unit of a digital moving image and compressed moving image data generated by lossless compression encoding of the digital moving image Decoding means for decoding the compressed moving image data to obtain a decoded moving image; signature calculating means for calculating a second signature based on the predetermined rule for each of the predetermined units of the decoded moving image; Detecting means for detecting whether or not the decoded moving image is normal based on the coincidence between the first signature and the second signature.

Thus, it is possible to confirm whether or not the losslessly compressed moving image data has been correctly decoded based only on the image representative information having a very small information amount regardless of the screen size.

In the present invention, it is preferable that the signature is a value calculated by inputting at least all bits of at least an effective pixel portion of a digital moving image or a decoded moving image to a linear feedback register.

[0032]

Embodiments of the present invention will be described below with reference to the drawings. The present invention refers to or processes a signature as a representative value of the original image or an image (for example, a decoded image) of the same size as the original image in moving image processing such as analysis, search, and collation of the moving image. That's it.

In the following embodiments, a case where the target moving image is an HDTV will be described as an example. In the present invention, transmission includes not only transmission in a narrow sense through a wired or wireless communication path, but also recording / reproduction or, in some cases, physical transportation of a recording medium.

(First Embodiment) First, a first embodiment of the present invention will be described. FIG. 1 is a schematic configuration diagram of a moving image processing apparatus according to the first embodiment of the present invention.

The moving image processing apparatus shown in FIG.
01, signature calculation unit 102, signature storage unit 10
3. It comprises a moving image analysis unit 104, an analysis result storage unit 105, and a compression encoding unit 106. Not all of these components need to work at all times,
Some or all of them operate as needed.

The original image storage unit 101 is, for example, a non-compression recording type HDTV digital VTR.
An original TV image is stored. From the HDTV original image reproduced from the original image storage unit 101, the signature calculation unit 1
At 02, the signature for each screen is sequentially calculated as described later, and the obtained list of signatures is stored in the signature storage unit 103. Generally, the signature calculation only needs to be performed once at the beginning. The signature stored in the signature storage unit 103 is stored in the moving image analysis unit 10.
4, where the analysis of the moving image using the signature is performed as described later. This analysis result is stored in the analysis result storage unit 105 as needed.

On the other hand, if necessary, the original HDTV original image is reproduced from the original image storage unit 101 and
06 to be compressed by, for example, the MPEG2 system. At the time of encoding by the compression encoding unit 106, the moving image analysis result stored in the analysis result storage unit 105 is referred to if necessary. In the present embodiment,
It is not indispensable to perform compression encoding, and therefore, some of the connections in FIG. 1 are indicated by dotted lines.

Hereinafter, the present embodiment will be described in more detail. First, a signature technique employed as one element of the present invention will be described. This signature technology is described in, for example, a literature (“Failure Checking Method for Microprocessor Application System”, Chapter 7, GB Willi).
ams, translated by Nobuyuki Goto, published by Japan Technology Economic Center Co., Ltd.). The basic concept of the signature is as follows.

A known test pattern is applied to the digital system, and a "signature" is calculated from a bit pattern passing through a certain node according to a predetermined rule. This signature reflects the information of all the bit patterns passing through the node, and should take a unique value depending on the test pattern and the node. Therefore,
By comparing this signature with an expected value determined in advance, the presence or absence of a fault from the input of the digital system to the node can be checked. That is, the signature is unique to the bit pattern and can be used to identify the bit pattern.

In the present invention, the signature is applied to a digital moving image, and the signature is obtained in units of one field or one frame image. That is, the signature plays a role as a representative value of an image of one field or one frame.

For the calculation of the signature, for example, a linear feedback shift register as used for generating an M sequence (maximum long period sequence) is used. The above literature includes a linear feedback shift register (LFSR; Linear Feedba).
1 shows an example of a signature calculation circuit that obtains a 16-bit signature from serial data using ck Shift Register.

FIG. 2 is a configuration diagram of the signature calculation unit 102 in the present embodiment. This is an improvement of the signature calculation circuit described in the above document so that it can be applied to parallel data. 2 includes an 8-bit register (D) 201, a parallel / serial conversion circuit (P / S) 202, and a seven-stage register (7).
D) 203, two-stage register (2D) 204, three-stage register (3D) 205, four-stage register (4D) 206, EX
It is composed of OR circuits 207, 208, 209 and 210.

7-stage register (7D) 203, 2-stage register (2D) 204, 3-stage register (3D) 205, 4-stage register (4D) 206, EXOR circuits 207 and 20
The portion consisting of 8, 209 and 210 constitutes a linear feedback shift register. The positions of the taps of the linear feedback shift register are determined so that a sequence having the maximum period that can be generated by a 16-bit shift register when input data is fixed at 0 is considered. Of course, there is a way of selecting taps that satisfies this condition other than that shown in FIG.

The image data (8
Is input in synchronization with the clock signal CLK, and is delayed by one clock (D) by the clock signal CLK in the 8-bit register (D) 201. The parallel / serial conversion circuit (P / S) 202 first holds a total of 16 bits of the input image data and the output of the 8-bit register 201 with a clock ク ロ ッ ク CLK having a frequency which is 倍 times the frequency of the CLK, and provides a 16: 1 parallel clock. After serial conversion, C
A 1-bit width serial sequence synchronized with 8CLK having a frequency eight times LK is output. 7-stage register (7D) 20
3, two-stage register (2D) 204, three-stage register (3
D) 205 and the four-stage register (4D) 206 shift data in synchronization with 8CLK, respectively. The EXOR circuits 207, 208, 209, and 210 take the EXOR of the two bits that are input to each, and output the result.

First, before starting calculation of the signature of one image (one field or one frame), a seven-stage register (7D) 203, a two-stage register (2D) 204, a three-stage register (3D) 205, a four-stage register Register (4D) 206
Are initialized to predetermined values (for example, all 1). In this state, input of image data (effective pixel portion) to be subjected to signature calculation to the linear feedback shift register is started. When all the data to be calculated have been input, the last value remaining in the shift register is taken out to obtain a 16-bit signature for the image.

Although the signature calculation method is not limited to the above, even if another calculation method is used, all bits of the effective pixel portion in one field (or one frame) of the digital moving image are reflected. Needless to say, such an arithmetic method should be used.

Also, regarding the number of bits of the signature,
Not only 16 bits that can be generated by the circuit of FIG. 2 but also 32 bits, 64 bits, and the like can be selected. Of course, the larger the number of bits, the lower the possibility of erroneous determination that the same signature is generated from different bit patterns. For example, assuming that one signature per field is obtained, assuming that signature calculation is continuously performed for an uninterrupted random input image, the frequency at which the same signature value is observed is determined by the 16-bit signature. About once every 18 minutes, about once every 2.3 years for a 32-bit signature, and about once every 10th year for a 64-bit signature.

This is because the target moving image is the same as the HDTV or the standard TV, and it is not necessary to increase the number of bits of the signature just because it is an HDTV. in this way,
Regardless of the screen size, information representing an image can be generated with a very small amount of information.

FIG. 3 is a diagram showing an example of the order of inputting image data to the signature calculation unit 102. HDTV 1
The number of effective pixels in the field is such that the luminance signal Y is horizontal 1920
× vertical 540, color difference signals Pb and Pr are horizontal 96
It is assumed that 0 × vertical 540 and the number of quantization bits is 8 bits. In this example, for an image of one field, first, Y
, The pixels are input in the raster scan order for Pb, and finally the pixels are input in the raster scan order for Pr. This input order is an example, and for example, Pb-Y-
The luminance signal and the chrominance signal may be multiplexed on a pixel-by-pixel basis, such as Pr, Y, Pb, Y, and then input to the signature calculation unit 102. In short, when calculating signatures, image data must be input in the same order.

As described above, in calculating the signature, the connection of the signature calculation circuit, the value of the initial value set in the register in the signature calculation circuit, the range of the image to be calculated, and the input order of the image data are always fixed. I do.

As described above, in the present embodiment, the original image is not transmitted, or is transmitted separately through a transmission path different from the signature. Naturally, the original image storage unit 101 requires an enormous storage capacity, and its transmission requires an extremely high transmission speed.

According to the present invention, in the moving image analysis unit 104,
The feature is that only the signature is used without using the original image. As can be understood from the above description,
Since the information amount of the signature is extremely small compared to the information amount of the original image, the storage capacity of the signature storage unit 103 may be extremely small, and the transmission speed thereof may be very low.
Also, in the processing in the moving image analysis unit 104, high-speed dedicated hardware or the like is not required, and high-speed processing can be performed by a computer program using a general-purpose processor unit usually built in a personal computer or the like.

Next, the moving picture analyzing unit 104 will be described in more detail. In the present embodiment, the moving image analysis unit 104
Is further subdivided into three and performs three types of processing. That is, there are three parts: a 3: 2 pull-down pattern detection unit, a still image part detection unit, and an image search unit.

(1) Detection of 3: 2 Pulldown Pattern First, a case where a 3: 2 pulldown pattern is detected using a signature will be described. The 3: 2 pulldown refers to converting a film image of 24 frames per second into an interlaced image of 60 fields per second by a telecine device or the like. This conversion is realized by repeating the image of the immediately preceding field of the same polarity (top or bottom) once every five fields. It is necessary to detect a 3: 2 pull-down pattern in order to restore the correspondence with the original film frame from the 3: 2 pull-down video. The 3: 2 pull-down pattern detection method in the present embodiment uses a repetition field (repeat field).
Can be applied digitally and repeatedly without any one-bit difference.

FIG. 4 is a diagram showing the configuration of the 3: 2 pull-down pattern detection unit. The 3: 2 pull-down pattern detection unit in FIG.
3, and 3: 2 pull-down determination unit 404.

The signatures for each field input from the signature storage unit 103 of FIG. 1 are sequentially input to the registers 401 and 402. The input signature and the signature two fields before, which are output from the registers 401 and 402, are supplied to the match detection unit 403, and it is detected whether or not the signatures match. This means that signature matching is monitored at one-frame intervals. The result of the coincidence detection is 3: 2 pull-down determination unit 404.
And a 3: 2 pull-down pattern is required.
At this time, the fact that the signature at one frame interval matches once every five fields is used.

FIG. 5 shows how a 3: 2 pull-down pattern is detected. In the range of this figure, fields 5 and 10 are repetitive fields (R) of fields 3 and 8, respectively. In the match detection unit 403, the output of the register 402 is
When the input of the register 401 is the field 5,
When the output of the register 402 is the field 8 and the register 401
Is input in the field 10, a signature match is detected (the match is indicated by a circle in FIG. 5). Otherwise, no signature match is detected (FIG. 5).
In this case, it is indicated by a cross).

Based on the result, the 3: 2 pull-down determination unit 404 restores the 3: 2 pull-down structure. That is, if the frame of the original film is expressed as a picture, the correspondence between the fields 1 to 12 and the pictures 1 to 5 as shown in FIG.

The 3: 2 pull-down determination unit 404 performs the determination on the assumption that there is a discontinuous point in the 3: 2 pull-down pattern due to editing or the like. In particular, when a 3: 2 pull-down pattern is detected by software as preprocessing, detection is also performed in the reverse direction in time, and the result is also referred to, particularly when there is a discontinuous point of the 3: 2 pull-down pattern. , The probability of correct detection is improved.

As described above, by adopting the image analysis processing method in which the signature is calculated from the original image and the image analysis is performed using the signature, the signature is obtained based on only the signature without referring to the original image. : 2 pull-down pattern can be detected.

The 3: 2 pull-down pattern detection result is useful alone, but may be used in the compression encoder 106 and the like. By providing the 3: 2 pull-down pattern to the compression encoding unit 106, it is possible to perform encoding after deleting repeated fields in advance. The original is a 24 frame film,
Since it is known that the image has a frame structure instead of a field structure, the coding efficiency can be improved.

(2) Detection of Still Image Portion Next, a case where a still image portion is detected by using a signature will be described. The still image here refers to a case where the screen is completely still without any digital difference of 1 bit.

FIG. 6 is a diagram showing the configuration of the still image portion detecting section. The still image portion detection unit in FIG. 6 has substantially the same configuration as the 3: 2 pull-down pattern detection unit in FIG.
The difference is that the 3: 2 pull-down determination unit 404 is replaced with the still part determination unit 601.

Also in this case, it is assumed that the signature is calculated in units of one field, and the coincidence detecting unit 403 monitors the coincidence of the signature at one frame interval. The result of the coincidence detection is supplied to the still portion judging section 601 to obtain a still image portion.

FIG. 7 is a diagram showing how a still image portion is detected. In the match detection unit 403, when the input of the register 401 is field 7, field 8, field 9, and field 10 and the output of the register 402 is field 5, field 6, field 7, and field 8, Is detected (see FIG. 7).
In this case, it is indicated by a circle). Otherwise, no signature match is detected (indicated by a cross in FIG. 7). Based on this result, the still part determination unit 601 determines that three frames of fields 5.6, 7.8, and 9/10 are still image parts as shown in FIG.

Note that the still image portion may be detected by using the signature for each frame. In this case, the configuration is basically the same as that in FIG. 6 except that the register 402 is deleted.

As described above, by adopting an image analysis processing method of calculating a signature from an original image and performing image analysis using the signature, a still image can be obtained based on only the signature without referring to the original image. Image portions can be detected.

Although the still image portion detection result is useful by itself, it may be used in the compression encoder 106 and the like. For example, by providing information on a still image portion to the compression encoding unit 106, detection of a motion vector and encoding itself can be suspended, and if encoding is performed by hardware, power consumption can be reduced by encoding by software. Then, the processing speed can be improved. In addition, depending on the encoding method, by performing encoding that sends only information indicating that the image is stationary, the code amount can be significantly reduced.

The register 40 is provided between the 3: 2 pull-down pattern detecting section in FIG. 4 and the still picture portion detecting section in FIG.
1, 402 and the coincidence detection unit 403 can be shared.

(3) Image Retrieval Further, a case in which an image is retrieved using a signature will be described. FIG. 8 is a diagram illustrating a configuration of the image search unit. The image search unit in FIG. 8 includes a register 801, a search signature setting unit 802, a specific signature storage unit 803, a match detection unit 804, and a search determination unit 805.

The signatures in units of one field or one frame input from the signature storage unit 103 of FIG. 1 are sequentially input to the register 801. Register 80
The output of No. 1 is supplied to the match detection unit 804, and it is detected whether or not the output matches the signature preset in the search signature setting unit 802. This match detection result is supplied to the search determination unit 805, and the search result is obtained. The specific signature storage unit 803 stores a specific signature described later in advance.

The search signature setting unit 802 may manually input the signature value of the search target screen, or may load a specific signature from the specific signature storage unit 803. Further, the input signature can be directly stored in the specific signature storage unit 803. For example,
When a signature is calculated while a certain original image is being reproduced and the cue point button is pressed, the signature of the image at that point is stored in the specific signature storage unit 803. This makes it possible to easily search for a screen that matches a certain playback screen.

In this embodiment, a search for a specific screen can be performed. Here, the specific screen is, for example, a test signal such as a black screen (black burst) or a white screen, or a standard color bar. Or,
The image may have a special meaning for a certain user.
Signatures corresponding to these specific screens are calculated in advance as specific signatures and are stored in the specific signature storage unit 8.
03, and these specific screens are searched by the match detection unit 804 matching this specific signature.

As described above, by adopting the image analysis processing method of calculating a signature from an original image and performing image analysis using the signature, an image can be obtained based on only the signature without referring to the original image. Can be searched.

Although the image search result is useful alone, it may be used in the compression encoding unit 106 and other devices. As an example, a case is considered where the signature is used as a trigger signal for starting and ending encoding in the compression encoding unit 106.

FIG. 9 is a diagram showing an example of a trigger method using a signature. FIG. 9A shows a case where the start and end points of the encoding are perceived by humans, and FIG. 9B shows a case where the start and end points of the encoding are not perceptible to humans. .

In FIG. 9A, encoding starts at point A,
It is assumed that encoding is to be terminated at point B. At this time, specific images are added before and after the encoding target image in a predetermined order. That is, as shown in FIG. 9A, a color bar is added for 10 seconds and a black burst for 10 seconds before point A, and a black burst for 10 seconds and a color bar is added for 10 seconds after point B. . The signature for the color bar (one frame) and the signature for the black burst (one frame) are calculated in advance and are assumed to be x and y, respectively. These are stored in the specific signature storage unit 803 in advance, and are loaded in the search signature setting unit 802.

In the search judging unit 805, the signature x is 1
If the signature y continues for 0 second and is detected for 10 seconds, it is determined to be point A, and encoding is started from the frame immediately after. Similarly, when the signature y continues for 10 seconds and the signature x continues for 10 seconds, the immediately preceding point is determined to be the point B, and the encoding is completed retroactively. In this case, it can be understood that the coding is the start / end of the coding even when the original picture is visually observed.

In FIG. 9B, encoding starts at point C,
Suppose that it is desired to end encoding at point D. At this time, a specific image is also added before and after the encoding target image. That is, as shown in FIG. 9 (b), one black burst 1 and one black burst 2 frame before point C, one frame of black burst 3 and one frame of black burst 4 after point D, respectively. Has been added. The signatures for these four types of black burst signals (one frame each) are calculated in advance,
These are S1, S2, S3, and S4, respectively. These are stored in the specific signature storage unit 803 in advance,
The search signature is set in the search signature setting unit 802.

When the search determining unit 805 detects the signature S1 for one frame and the signature S2 for one frame,
This is determined as point C, and coding is started from the frame immediately after. Similarly, when the signature S3 is detected for one frame and the signature S4 is detected for one frame, the point immediately before that is determined to be the point D, and the encoding is completed retroactively.

Here, the four types of black burst signals have different LSBs of certain color difference signal pixels. That is, black burst 1 is upper left, black burst 2 is upper right, black burst 3 is lower left, black burst 4
In the figure, the LSB of the color difference signal pixel at the lower right corner of the screen is inverted. Except for this, the image is exactly the same as a normal black burst signal. In this case, since only the LSB is different, the difference between them is not apparent just by visual inspection, but the signatures are completely different.

That is, in the case of FIG.
Utilizing that the signature has a completely different value even with a difference in the bits, information for coding start / end control is embedded in the signature so as to be invisible to the naked eye.

Of course, the application of the control method using such a signature is not limited to the start and end of encoding. Further, the start / end of the encoding may be controlled by using the signature of the original original image without intentionally inserting the specific screen.

Also, for example, a signature can be used in place of a time code to specify an editing in-point, out-point, or the like by a signature. If not only a single signature but also all signature sequences after the pre-roll point are specified, it can be used for synchronization control when it takes a long time to establish synchronization such as in a VTR. However,
If the still image continues, the signature values will all be the same, so it is better to separately calculate a signature that is a function of time and use this together with the search.

As described above, according to the present embodiment, the processing in the moving image analysis unit 104 can be greatly simplified. Therefore, even if the processing is not performed by hardware, a computer program including a procedure corresponding to the operation of the moving image processing apparatus according to the first embodiment can be processed by software at a sufficiently high speed. In this case, 3: 2 pulldown pattern detection, still image portion detection,
In addition to analysis processing such as image search, if necessary,
A part or all of the signature calculation, the encoding process, and the decoding process may be performed by a computer program. By mounting such a computer program on a computer via a recording medium, or by introducing it to a general-purpose computer via a recording medium and executing it, the same effects as in the present embodiment can be obtained.

The moving image that can be handled in the present invention is limited to a digital moving image having no noise or error, and it is necessary that the moving image does not go through an analog process. It is also necessary that no errors occur in the transmission of digital moving images. Otherwise, the reproducibility of the signature is not guaranteed.
Conversely, by using the signature, it is possible to detect whether an analog process has been performed or an error has occurred.

Up to this point, it has been assumed that the signature is calculated in units of one field or one frame. However, the signature of the entire moving image or the signature for one second may be calculated. In this case, after initialization is performed at the beginning of each calculation unit, data is continuously input to the signature calculation circuit for the target period, and the last value remaining in the register may be used as the signature.

Further, not only the signature but also the time code, the DC average value of each of the luminance signal and the color difference signal, and the like may be combined as the image representative value. The information amount of the image representative value is extremely large even if these are included. By referring to these at the same time, the probability of erroneous detection of the image can be reduced and the characteristics of the image can be represented.

As described above, according to the present embodiment, the signature is calculated in units of one field or one frame of the original image, and the analysis / search of the moving image is performed using only this signature. Therefore, the following effects can be obtained.

(1) There is no need to refer to the original image or an image of the same size as the original image. (2) Information representing an image can be generated with an extremely small amount of information. (3) A moving image can be analyzed and searched with reference to only the signature, and high-speed hardware or the like is not required, and processing can be performed by software.

(Second Embodiment) Next, a second embodiment of the present invention will be described. In this embodiment, a signature is transmitted together with compressed data using a rewritable optical disk as a transmission medium.

FIG. 10 is a schematic configuration diagram of a moving image processing apparatus according to this embodiment. The moving image processing apparatus in FIG. 10 includes an original image storage unit 1001, a compression encoding unit 1002, an original image signature calculation unit 1003, a local decoded image signature calculation unit 1004, a recording unit 1005, a recording medium 1006,
It comprises a reproducing unit 1007, a decoding unit 1008, a decoded image signature calculation unit 1009, a match detection unit 1010, a collation image signature calculation unit 1011, and a match detection unit 1012. Not all of these components need to operate at all times, and some or all of them may operate as needed.

HD reproduced from original image storage section 1001
The TV original image is supplied to a compression encoding unit 1002, where the original image is subjected to compression encoding according to, for example, the MPEG2 method, and compressed data is supplied to a recording unit 1005.

On the other hand, from the same HDTV original image, the signature for each screen is sequentially calculated in the original image signature calculation unit 1003, and the obtained original image signature is supplied to the recording unit 1005. A local decoded image for predictive encoding is output from the compression encoding unit 1002, and a signature for each screen is sequentially calculated in a local decoded image signature calculation unit 1004, and the obtained local decoded image signature is recorded in a recording unit. 1005. The local decoded image is used as a reference screen for a motion vector search or the like of the input screen,
When encoding is performed in the order of DCT, quantization, and variable-length encoding, the quantized DCT coefficient data is subjected to inverse quantization, inverse DCT (IDC
T), a local decoded image is obtained.

The recording unit 1005 multiplexes the compressed data, the original image signature, and the local decoded image signature and records them on the recording medium 1006. Here, the recording medium 1006
Is a rewritable optical disk.

The reproducing unit 1007 reproduces the compressed data, the original image signature, and the local decoded image signature from the recording medium 1006, and supplies them to the decoding unit 1008, the coincidence detecting unit 1012, and the coincidence detecting unit 1010, respectively. The decoding unit 1008 decodes the moving image from the compressed data, outputs a reproduced moving image, supplies the reproduced moving image to a display unit (not shown), and displays the reproduced image. The decoded data output from the decoding unit 1008 is supplied to the decoded image signature calculation unit 1009, the signature of each screen is sequentially calculated, and the obtained decoded image signature is supplied to the coincidence detection unit 1010.

Further, the collation image input from the outside is supplied to the collation image signature calculation unit 1011, the signature of each screen is sequentially calculated, and the obtained collation image signature is supplied to the coincidence detection unit 1012. The match detection unit 1010 detects a match between the reproduced local decoded image signature and the decoded image signature calculated by the decoded image signature calculation unit 1009. Also, the match detection unit 1012 detects a match between the reproduced original image signature and the collated image signature.

The input order of the image data to the original image signature calculation unit 1003 and the collation image signature calculation unit 1011 is the same as that in FIG. These signatures are calculated for each field or frame.

FIG. 11 shows a local decoded image signature calculating section 1004 or a decoded image signature calculating section 1009.
FIG. 6 is a diagram illustrating an example of an order in which image data is input to a printer. For these images, in this embodiment, the signature is calculated for each coded picture.

It is assumed that one picture of an HDTV is encoded by 120 horizontal macroblocks × 68 vertical macroblocks. Each macro block has four luminance signals Y,
Each of the color difference signals Pb and Pr is composed of a total of six DCT blocks of 8 × 8 pixels, and the number of quantization bits is eight bits. In this example, as shown in FIG. 11A, pixels of one picture are input in raster scan order in macroblock units. In each macroblock, as shown in FIG. 11B, pixels are first input for the luminance signal Y in the raster scan order for each DCT block, and then the pixels are input for the color difference signal Pb in the raster scan order. Finally, pixels are input in the raster scan order for the color difference signal Pr. This input order is an example, and another order may be used. In short, when calculating signatures, image data must be input in the same order.

(1) Collation of Original Image First, a case where image collation is performed in this embodiment, that is, a case where collation is performed to determine whether or not the image is a genuine original image will be described.

The input to the collation image signature calculator 1011 may be a correct original image or an incorrect or false original image. By detecting a match between the reproduced original image signature and the collation image signature, the coincidence detection unit 1012 can confirm whether the collation image is the original image. Match detection unit 1012
In the case where a mismatch of signatures is detected, a warning indicating that the collation image is different from the original image is displayed on a warning display unit (not shown).

In this embodiment, both the compressed data and the signature are recorded and reproduced on the recording medium 1006. Although the human cannot judge and understand the content of the image only by looking at the signature, the human can grasp the content of the video by decoding and displaying the compressed data.
On the other hand, inside the apparatus, the original image can be grasped and compared with a small amount of information called a signature.

As an application of the present embodiment, for example, a system for registering an original image with any registered certification authority is considered. The original image registration applicant brings the original image to the registered certification authority, calculates the original image signature, and compresses and encodes to generate compressed data. Then, the compressed data and the original image signature are recorded on the disk. here,
The calculation method of the original image signature is a secret in the registered certification authority. The registered certification authority keeps only the disk on which the compressed data and the original image signature are recorded. The applicant for registration keeps the original at his own risk. The registered certification authority can play back the compressed data from the disc and decrypt it, so that the contents of the video can be roughly confirmed.
Only when strict matching is required, the signature is checked by bringing the original applicant to the registration applicant. As described above, the registered certification institution does not need to manage and store the original image itself, so that the storage capacity can be saved.

(2) Detection of Normal Decoding Next, a case will be described in which the present embodiment checks whether or not decoding has been correctly performed.

The coincidence detecting section 1010 confirms whether or not decoding has been performed correctly by detecting a match between the reproduced local decoded image signature and the decoded decoded image signature. If the signature mismatch is detected, a warning indicating that decoding is not correctly performed is displayed on a warning display unit (not shown).

For example, DCT in MPEG2
In the case of the IDCT operation, a mismatch in the operation result may occur due to a difference in the operation algorithm (including the order) between the encoding side and the decoding side during the IDCT operation. In this case, it is not possible to check whether the decoding is normal because a slight error occurs in the decoding result, but on the contrary, it should be used to check whether an IDCT mismatch has occurred due to the signature match detection described above. Can be. That is, during encoding and decoding, I
It is possible to detect whether there is a difference between the DCT calculation algorithms.

As described above, in this embodiment, the encoding side records the compressed data, the original image signature, and the local decoded image signature on the same recording medium and transmits them, and the decoding side transmits the collation image signature. By using the method of calculating and comparing it with the received original image signature, or calculating the decoded image signature and comparing it with the received local decoded image signature,
It is possible to easily confirm whether or not the image to be verified is exactly the same as the original original image and whether or not decoding has been correctly performed.

In this embodiment, since the compressed data is transmitted together, the content of the image can be grasped by decoding and displaying the compressed data. The second
By preparing a computer program including a procedure corresponding to the operation of the moving image processing device according to the embodiment, the matching and the normal decoding detection processing may be performed by software.

(Third Embodiment) Next, a third embodiment of the present invention will be described. In this embodiment, compression encoding is performed by a lossless compression method, and the signature of the original image is transmitted together with the compressed data.

FIG. 12 is a schematic configuration diagram of a moving image processing apparatus according to this embodiment. The moving image processing apparatus in FIG. 12 includes an original image storage unit 1201, a compression encoding unit 1202, an original image signature calculation unit 1203, a transmission unit 1204, and a decoding unit 12.
05, a decoded image signature calculation unit 1206, and a match detection unit 1207.

The HD reproduced from the original image storage unit 1201
The TV original image is supplied to a compression encoding unit 1202, where lossless compression encoding is performed, and compressed data is output to the transmission unit 1
204. On the other hand, from the same HDTV original image,
The original image signature calculation unit 1203 sequentially calculates the signature for each screen, and supplies the obtained original image signature to the transmission unit 1204. The compressed data and the original image signature are multiplexed and transmitted by the transmission unit 1204.

The transmission section 1204 outputs the compressed data and the original image signature, and supplies them to the decoding section 1205 and the coincidence detecting section 1207, respectively. The decoding unit 1205 decodes the moving image from the compressed data, and outputs a reproduced moving image. The decoded data output from the decoding unit 1205 is supplied to a decoded image signature calculation unit 1206, the signature of each screen is sequentially calculated, and the obtained decoded image signature is supplied to the coincidence detection unit 1207. The match detection unit 1207 detects a match between the received original image signature and the received decoded image signature.

In the present embodiment, the signatures in the original image signature calculation unit 1203 and the decoded image signature calculation unit 1206 are calculated for each field or frame, and the input order is as shown in FIG.

In the above-described second embodiment, it is assumed that the compression coding method in the compression coding section is an irreversible compression method such as MPEG2. However, in this embodiment, the compression coding method is a lossless compression method. Has become.

A compression method such as MPEG2 is irreversible compression and cannot completely restore the original information, but can achieve a large compression rate instead. On the other hand, in the case of reversible compression, the original image can be completely restored, but the compression ratio cannot be so high.

In the present embodiment, the original image signature is calculated in advance and transmitted together with the lossless compressed data, so that the decoded image signature calculated after decoding is detected as coincident with the received original image signature. This makes it possible to confirm whether or not the lossless compression has been correctly decoded. When the coincidence detecting unit 1207 detects a mismatch between these signatures, a warning indicating that decoding is not correctly performed is displayed on a warning display unit (not shown).

As described above, in the present embodiment, when the compression encoding method is the lossless compression method, the signature of the original image is transmitted together with the compressed data, so that the signature of the received original image is used. Thus, it is possible to verify whether or not decoding has been correctly performed on the decoding side.

(Modification) The present invention is not limited to the above-described first to third embodiments, but may be, for example, the following (1) to (5).
Various modifications and applications are possible as exemplified in FIG.

(1) In the first to third embodiments, the target moving image is an HDTV, but may be a standard TV or a UDTV. (2) In the first to third embodiments, the signature is calculated using the linear feedback register. However, the present invention is not limited to this. Also, the calculation of the signature can be performed using a computer program.

(3) In the second embodiment, the recording medium is a rewritable optical disk. However, other recording media such as a magnetic tape, a magnetic disk, and a semiconductor memory may be used. In addition, even if it is not a recording medium,
A wireless communication path may be used.

(4) Not only one kind of signature but also a signature in field units and a signature in frame units may be obtained. Alternatively, one signature may be obtained from all the images and used as representative information of the entire image, or a signature may be obtained in units of one GOP image. Conversely, signatures may be calculated in units smaller than one field.

(5) As the image representative information, not only the signature but also other information may be included. Further, the time code itself may be included in the image representative information, or the time code may be included in the signature calculation. Also, for example, a signature in GOP units of the compressed data itself may be obtained and transmitted together.

[0124]

As described above, according to the present invention,
A moving image processing apparatus capable of performing moving image processing such as analysis, search, and collation of a moving image without referring to or processing the original image or an image of the same size as the original image can be provided. In particular, it is possible to provide a moving image processing apparatus that does not need to prepare a high-speed transmission path and dedicated processing hardware even for processing of a moving image with a high data rate such as HDTV or UDTV.

[Brief description of the drawings]

FIG. 1 is a block diagram showing a configuration of a moving image processing device according to a first embodiment of the present invention.

FIG. 2 is a block diagram showing a configuration of a signature calculation unit provided in the moving image processing apparatus according to the first embodiment.

FIG. 3 is an exemplary view for explaining the order of signature calculation in the moving image processing apparatus according to the first embodiment;

FIG. 4 is a block diagram showing a configuration of a 3: 2 pull-down pattern detection unit provided in the moving image processing device of the first embodiment.

FIG. 5 shows 3 in the moving image processing apparatus of the first embodiment.
FIG. 9 is a diagram for explaining the operation of a two pull-down pattern detection process.

FIG. 6 is a block diagram illustrating a configuration of a still image portion detection unit provided in the moving image processing device according to the first embodiment.

FIG. 7 is an exemplary view for explaining the operation of still image portion detection processing in the moving image processing device of the first embodiment.

FIG. 8 is a block diagram showing a configuration of an image search unit provided in the moving image processing device of the first embodiment.

FIG. 9 is an exemplary view for explaining an example of a trigger method applied to the moving image processing apparatus according to the first embodiment.

FIG. 10 is a block diagram showing a configuration of a moving image processing device according to a second embodiment of the present invention.

FIG. 11 is an exemplary view for explaining a calculation order of decoded image signatures in the moving image processing apparatus according to the second embodiment.

FIG. 12 is a block diagram showing a configuration of a moving image processing device according to a third embodiment of the present invention.

[Explanation of symbols]

 Reference Signs List 101: Original image storage unit 102: Signature calculation unit 103: Signature storage unit 104: Moving image analysis unit 105: Analysis result storage unit 106: Compression coding unit 201: 8-bit register 202: Parallel-serial conversion circuit 203: 7-stage register 204 ... two-stage register 205 ... three-stage register 206 ... four-stage register 207, 208, 209, 210 ... EXOR circuit 401, 402 ... register 403 ... match detection unit 404 ... 3: 2 pull-down judgment unit 601 ... stationary part judgment unit 801 ... Register 802 ... Search signature setting unit 803 ... Specific signature storage unit 804 ... Match detection unit 805 ... Search determination unit 1001 ... Original image storage unit 1002 ... Compression coding unit 1003 ... Original image signature calculation unit 1004 ... Local decoded image signature calculation Unit 1005 ... Recording unit 1 06 Recording medium 1007 Reproduction unit 1008 Decoding unit 1009 Decoded image signature calculation unit 1010 Match detection unit 1011 Matching image signature calculation unit 1012 Match detection unit 1201 Original image storage unit 1202 Compression encoding unit 1203 Original image signature calculation unit 1204 ... Transmission unit 1205 ... Decoding unit 1206 ... Decoded image signature calculation unit 1207 ... Match detection unit

──────────────────────────────────────────────────の Continued on front page (51) Int.Cl. ⁶ Identification code FI H04N 7/18 H04N 7/13 Z

Claims (12)

[Claims] 1. A signature calculating means for calculating a signature based on a predetermined rule for each predetermined unit of a digital moving image, and based on image representative information including at least the signature, without referring to the digital moving image. A moving image processing apparatus comprising: an analyzing unit for analyzing a digital moving image. 2. Signature calculating means for calculating a signature based on a predetermined rule for each predetermined unit of a digital moving image, and transmitting means for transmitting image representative information including at least the signature separately from the digital moving image. A moving image processing apparatus comprising: 3. Receiving means for receiving image representative information including at least a signature calculated based on a predetermined rule for each predetermined unit of a digital moving image, and based on the image representative information received by the receiving means, Analyzing means for analyzing the digital moving image without referring to the digital moving image. 4. Image representative information including at least a signature calculated based on a predetermined rule for each field of the digital moving image is input, and the digital moving image is converted based on a matching state of one frame interval of the signature. A moving image processing apparatus, comprising: detecting means for detecting a 3: 2 pull-down pattern of the digital moving image without reference. 5. Image representative information including at least a signature calculated based on a predetermined rule for each field or frame of a digital moving image is input, and based on a matching state of one frame interval of the signature, the digital moving image is displayed. Detecting means for detecting whether the digital moving image is completely still without referring to an image. 6. Image representative information including at least a first signature calculated based on a predetermined rule for each field or frame of a digital moving image is input, and said first signature and a second preset image are set. A moving image processing apparatus comprising: a search unit that searches for the digital moving image without referring to the digital moving image based on a matching state of the signature. 7. A signature calculating means for calculating a signature based on a predetermined rule for each predetermined unit of the digital video, a compression means for compressing and coding the digital video to generate compressed video data, and the signature And a transmitting unit for transmitting image representative information including at least the compressed moving image data together with the compressed moving image data. 8. A method for compressing and encoding image representative information including at least a first signature calculated based on a predetermined rule for each predetermined unit of a first digital moving image and the first digital moving image, and generating the first digital moving image. Receiving means for receiving the compressed moving image data, decoding means for decoding the compressed moving image data to obtain a decoded moving image, and for each of the predetermined units of the second digital moving image, based on the predetermined rule. Signature calculating means for calculating a second signature; and wherein the second digital moving image is the same as the first digital moving image based on a matching state between the first signature and the second signature. A moving image processing apparatus, comprising: a detecting unit for detecting whether or not the moving image is detected. 9. A compression means for compressing and encoding a digital moving image to generate compressed moving image data, and a signature for calculating a signature based on a predetermined rule for each predetermined unit of a locally decoded moving image of the compressed moving image data. A moving image processing apparatus comprising: calculating means; and transmitting means for transmitting decoded image representative information including at least the signature together with the compressed moving image data. 10. A compressed moving image data generated by compressing and encoding a digital moving image, and a first signature calculated based on a predetermined rule for each predetermined unit of a locally decoded moving image of the compressed moving image data. Receiving means for receiving decoded image representative information including at least: a decoding means for decoding the compressed moving image data to obtain a decoded moving image; and for each of the predetermined units of the decoded moving image, based on the predetermined rule. Signature calculating means for calculating a second signature; and detecting means for detecting whether or not the decoded moving image is normal based on the matching status between the first signature and the second signature. A moving image processing device characterized by the above-mentioned. 11. Image representative information including at least a first signature calculated based on a predetermined rule for each predetermined unit of a digital moving image, and compressed moving image data generated by reversibly compressing and encoding the digital moving image. A decoding unit that decodes the compressed moving image data to obtain a decoded moving image; and a signature that calculates a second signature based on the predetermined rule for each of the predetermined units of the decoded moving image. A moving image processing apparatus comprising: calculating means; and detecting means for detecting whether or not the decoded moving image is normal based on the matching status between the first signature and the second signature. . 12. The signature according to claim 1, wherein the signature is a value calculated by inputting at least all bits of an effective pixel portion of a digital moving image or a decoded moving image to a linear feedback register. 3. The moving image processing device according to claim 1.