JP2002010254A - Feature point detection method and record reproduction device - Google Patents

Abstract

PROBLEM TO BE SOLVED: To make it possible to detect scene change points directly from coded visual signals without decoding coded visual data, thus reducing the amount of operation for detecting scene change points. SOLUTION: Coded video and sound data from external equipment are converted to TS in a digital interface circuit 90, and PES is extracted from the TS in a separation circuit 42, and scene change point is detected by a feature point detection part 100. In the feature point detection part 100, appearance period of I picture is detected, and if it becomes shorter than a predetermined period, then the I picture is set as a scene change point. If the data length of P picture or B picture becomes longer than threshold values LthP or LthB which are the result obtained by multiplying the coefficient by the mean value of P picture or B picture data length in the closest past plural frames stored in a memory 104, then the P picture or the B picture at that point is set as a scene change point.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a method for detecting a scene change point of an image from video data, and an apparatus for recording and reproducing video data.

[0002]

2. Description of the Related Art Video data is analyzed by a recording / reproducing apparatus, and a transition between a program and a program, a transition between a program and a commercial message portion, a point at which a video content in a program greatly changes, and the like are determined by scene change points of an image. Some are provided with a function of detecting as.

Such a recording / reproducing apparatus realizes random access reproduction such as cue reproduction, reproduction from the end of a commercial message, and reproduction from a specific scene in a program, based on the information of the detected scene change point. And thumbnails (reduced images) of scene change points can be easily created.

Conventionally, the detection of this scene change point has been performed by analyzing baseband video data before or after encoding. Specifically, the histogram of the signal intensity of the luminance component or the color difference component of the frame at that time is compared with the similar histogram of the immediately preceding frame, and the correlation between adjacent frames is detected. If there is no scene change and the correlation is weak, it is determined that a scene change has occurred.

[0005]

However, when a scene change point is detected by such a method, the amount of calculation for detecting the scene change point increases, and the load on the system controller increases.

In a conventional method of detecting a scene change point from baseband video data before or after encoding, encoded video data input from an external digital broadcast receiver or the like is used. When recording, the encoded video data must be decoded once to detect a scene change point, so when detecting a scene change point, the video decoder cannot be used for another purpose. However, encoded video data already recorded on a recording medium cannot be reproduced from a recording medium and decoded by a video decoder to obtain a reproduced video signal.

Accordingly, the present invention is capable of detecting a scene change point directly from encoded video data without decoding the encoded video data, and reducing the amount of calculation for detecting the scene change point. Can be done,
When a scene change point is detected, the video decoder can be used for another purpose.

[0008]

A feature point detection method according to a first aspect of the present invention is a method for detecting a scene change point of an image as a feature point from video data encoded as a plurality of types of pictures. The appearance cycle of a specific type of picture in the coded video data is detected, and when the appearance cycle becomes shorter than a predetermined period, the specific type of picture at that time is set as a scene change point.

A feature point detecting method according to a second aspect of the present invention is a method for detecting a scene change point of an image as a feature point from video data coded as a plurality of types of pictures. The data length of each picture other than the specific type is detected, and when the data length becomes larger than the threshold value, the other type of picture at that time is set as a scene change point.

According to the feature point detecting method of the first or second aspect of the present invention, by analyzing the picture structure of the encoded video data, the encoded video data can be directly converted from the encoded video data.
Since the scene change point is detected, the amount of calculation for detecting the scene change point can be reduced, and the load on the system controller can be reduced. Further, since a scene change point can be detected without decoding encoded video data, the video decoder can be used for another purpose when detecting a scene change point.

[0011]

DESCRIPTION OF THE PREFERRED EMBODIMENTS [Overview of Recording / Reproducing Apparatus and System ... FIGS. 1 and 2] FIG. 1 shows an example of a recording / reproducing apparatus according to the present invention.
(Moving Picture Experts G
loop), compression-encodes, multiplexes, and
PEG system TS (Transport Street)
am) as a HDD (Hard Disk Drive)
This is a case where the data is recorded on the disk (hard disk) in e) and reproduced from the disk.

Hereinafter, the analog video signal and the analog audio signal are referred to as a video signal and an audio signal, respectively.
The digital video signal and digital audio signal are referred to as digital video data and digital audio data, or simply video data and audio data.

In this example, the recording / reproducing apparatus 10 is connected to an antenna 1 for receiving a terrestrial TV (television) signal, and has analog input terminals 11V, 11A, 11A.
S, analog output terminals 12V, 12A, 12S, digital input / output terminal 19, input processing unit 20, output processing unit 30,
Demultiplexing processing unit 40, buffer control circuit 50, HDD 6
0, system controller 70, synchronization control circuit 80, digital interface circuit 90, and feature point detection unit 10
0 is provided. As shown in FIG. 2, the demultiplexing processing unit 40 includes a multiplexing circuit 41 and a demultiplexing circuit 42.

Although not shown in the figure, the system controller 70 has a CPU (Central Processing).
Unit), ROM (Read Only) in which a program to be executed by the CPU and fixed data are written.
Memory) and a RAM (Random Access Mem) functioning as a work area of the CPU and the like.
(ory), and controls the entire recording / reproducing apparatus 10.

(Overview of Input Processing and Recording) Antenna 1
The terrestrial TV signal received at is selected by the terrestrial tuner 21 to obtain a video signal and an audio signal from the terrestrial tuner 21. The video signal and the audio signal are supplied to one input terminal of the input switching circuit 22. Supplied.

The input terminal 11V is supplied with a composite video signal from an external device, the input terminal 11A is supplied with an audio signal from the external device, and the input terminal 11S is supplied with a separate video signal from the external device.

The composite video signal from the input terminal 11V and the audio signal from the input terminal 11A are supplied to the other input terminal of the input switching circuit 22.
Is switched by the system controller 70,
One of the video signal and the audio signal is selected and taken out from the input switching circuit 22.

The video signal from the input switching circuit 22 is YC
The signal is separated into a luminance signal and a color difference signal by a separating circuit 23 and supplied to one input terminal of another input switching circuit 24. Also,
A separate video signal (luminance signal and color difference signal) from the input terminal 11S is supplied to the other input terminal of the input switching circuit 24, and the input switching circuit 24
0, and any one of the luminance signal and the color difference signal is selected and taken out from the input switching circuit 24.

The luminance signal and the color difference signal from the input switching circuit 24 are output from an NTSC (National Televi).
The A / D conversion is performed by the S / N system decoder 25, and the image data is further subjected to chroma-encoding processing to obtain component video data from the NTSC decoder 25.

In the NTSC decoder 25, a vertical synchronizing signal and a horizontal synchronizing signal are separated from the luminance signal from the input switching circuit 24, and a clock and a field discrimination signal are generated based on the separated synchronizing signal. The synchronization signal, the clock, and the field discrimination signal are supplied to the synchronization control circuit 80. In the synchronization control circuit 80, the recording / reproducing device 10
A clock and a timing signal required for each section are generated and supplied to each section of the recording / reproducing apparatus 10.

The video data from the NTSC decoder 25 is subjected to video processing such as pre-filtering by a pre-video processing circuit 26, and then supplied to an MPEG video encoder 27 and a post-video processing circuit 32 of an output processing unit 30. .

In the MPEG video encoder 27, the video data from the pre-video processing circuit 26
T (Discrete Cosine Transfo
rm: Discrete Cosine Transform) and the like, and the video ES (Elementary Stream)
Is generated, and the MPEG video ES is supplied to the demultiplexing processing unit 40.

On the other hand, the audio signal from the input switching circuit 22 is converted into digital audio data by the audio A / D converter 28 and then supplied to the MPEG audio encoder 29 and the output switching circuit 35 of the output processing unit 30. .

In the MPEG audio encoder 29, audio A
The audio data from the / D converter 28 is compressed according to the MPEG format to generate an audio ES, and the MP
The EG sound ES is supplied to the demultiplexing processing unit 40.

In the demultiplexing processing section 40, the video ES from the MPEG video encoder 27 and the audio ES from the MPEG audio encoder 29 are multiplexed with various control signals in a multiplexing circuit 41 shown in FIG. Is generated, and the MPEG / TS is transmitted to the buffer control circuit 50 via the selector 44.

The buffer control circuit 50 includes a multiplexing circuit 41
, Which are continuously input from the selector via the selector 44
/ TS to the HDD 60 intermittently. That is,
When the HDD 60 is performing a seek operation, writing cannot be performed. Therefore, the input MPEG / TS is temporarily stored in a buffer, and when the HDD 60 is writable, the MPEG / TS is stored in the buffer at a higher rate than the input rate. The data is read out and sent to the HDD 60. Thus, M continuously input to the buffer control circuit 50
The PEG / TS is recorded on the disk (hard disk) 61 without interruption.

The HDD 60 includes a system controller 70
And writes the MPEG / TS to the disc 61. As a protocol (interface) between the buffer control circuit 50 and the HDD 60, IDE (In
graded Device Electroni
cs) is used.

(Overview of Reproduction and Output Processing) During reproduction, the HDD 60 is controlled by the system controller 70 to read MPEG / TS from the disk 61 and send it to the buffer control circuit 50. The buffer control circuit 50 converts the MPEG / TS intermittently input from the HDD 60 into a continuous MPEG / TS, contrary to the recording operation, and converts the MPEG / TS into a continuous MPEG / TS through the selector 46 of the demultiplexing processing unit 40. 42.

In the separation circuit 42, the continuous MPEG
/ TS to PES (Packetized Element)
ntary Stream) is extracted and supplied to the MPEG video / audio decoder 31.

In the MPEG video / audio decoder 31, the PES from the separation circuit 42 is separated into a video ES and an audio ES, and the video ES is decoded by an MPEG video decoder and converted into baseband video data.
S is decoded by the MPEG audio decoder and converted into baseband audio data. The converted video data is supplied to a post video processing circuit 32, and the audio data is supplied to an output switching circuit 35.

In the post video processing circuit 32, video processing such as switching or synthesizing or post filtering is performed on the video data from the MPEG video / audio decoder 31 and the video data from the pre-video processing circuit 26. The video data after the processing is OSD (On Sc)
(replay) processing circuit 33.

In the OSD processing circuit 33, the video data from the post video processing circuit 32 is subjected to processing such that an image such as graphics is superimposed or partially displayed on the display screen. The data is supplied to the NTSC encoder 34.

In the NTSC encoder 34, the video data (component video data) from the OSD processing circuit 33 is converted into luminance data and chrominance data, and the luminance data and chrominance data are respectively converted to D / D data.
The signal is A-converted to obtain a composite video signal and a separate video signal of an analog signal. The composite video signal is led to the output terminal 12V, and the separate video signal is led to the output terminal 12S.

On the other hand, in the output switching circuit 35, the MPEG video / audio decoder 3 is controlled by the system controller 70.
1 and the audio data from the audio A / D converter 28 are selected and taken out. The selected audio data is converted into an analog audio signal by the audio D / A converter 36, and is output to the output terminal 12A.

(Outline of Interface with External Device) In the recording / reproducing device 10, a digital interface circuit 90 is connected between the digital input / output terminal 19 and the demultiplexing processing unit 40, and the digital input / output terminal 19, the external device 110 is connected.
Can be recorded on the disk 61, and the encoded video and audio data reproduced from the disk 61 can be output to the external device 110.

As the external device 110, an IRD (Int
It is possible to connect video equipment such as an eGraded ReceiverDecoder and a digital TV receiver, or information equipment such as a personal computer. As a digital interface between the external apparatus 110 and the recording / reproducing apparatus 10, an IEEE (Institute)
of Electric and Electr
onics Engineers) 1394 standard interface or the like can be used.

Digital input / output terminal 1 from external device 110
9 is subjected to processing such as format conversion by the digital interface circuit 90, and the MPEG / T
It is converted to S and sent to the demultiplexing processing unit 40. The demultiplexing processing unit 40 analyzes and generates a control signal as necessary, sends the MPEG / TS to the buffer control circuit 50 via the selectors 45 and 44, and sends the MPEG / TS to the disk 61 by the HDD 60. Be recorded.

At the same time, the MP through the selector 45
The EG / TS is supplied to the separation circuit 42 via the selector 46, and the PES is extracted from the recorded MPEG / TS by the separation circuit 42, and the MPEG video / audio decoder 31
, An analog video signal and an analog audio signal based on the video and audio data from the external device 110 are obtained at the output terminals 12V, 12S, and 12A.

At the time of reproduction, MPEG / TS is read from the disk 61 by the HDD 60, converted into continuous MPEG / TS by the buffer control circuit 50, and sent to the demultiplexing processing section 40. In the demultiplexing processing unit 40,
After analyzing and generating the control signal as required, the MPEG / TS is sent to the digital interface circuit 90 via the selector 45. In the digital interface circuit 90, the MPEG / TS is converted into video / audio data suitable for the external device 110 by a process reverse to that at the time of recording, and output to the external device 110 via the digital input / output terminal 19.

At the same time, the reproduced MPEG / TS from the buffer control circuit 50 is supplied to the separation circuit 42 via the selector 46, and the separation circuit 42 extracts the PES from the reproduced MPEG / TS. , MPEG video / audio decoder 31 to obtain a reproduced analog video signal and a reproduced analog audio signal at output terminals 12V, 12S and 12A.

[Feature point detection method: FIGS. 3 to 12] In the recording / reproducing apparatus 10 as described above, the digital input / output terminal 19 is input from the external device 110 and the MPEG / TS A method of detecting a feature point (scene change point) when the video / audio data converted into the above is recorded on the disk 61 will be described below.

(Encoding Mode of Video Data: FIGS. 3 and 4) In the MPEG system, video data is I (Intra)
-Coded), P (Predictive-code)
d), B (Bidirectionally pred)
coded-coded), and in the coded video data, one I
A GOP (Group Of Pictures) structure composed of pictures and several P pictures and B pictures is formed, and generally, I pictures appear periodically.

However, some digital broadcast encoders suppress the deterioration of image quality by forcibly inserting an I-picture when the video content greatly changes.

FIG. 3 shows the G of the coded video data in this case.
4 shows the OP structure, and the regular appearance period N of the I picture is 1
5A shows a GOP structure when there is no scene change, and FIG. 12B shows an I picture inserted at a position indicated by an arrow due to a scene change. Is a GOP structure in a case where the appearance period of the is disturbed.

In the first method of feature point detection, a scene change point is detected by detecting the disorder of the appearance cycle of the I picture.

Some digital broadcast encoders do not forcibly insert an I picture even if the video content greatly changes, and continuously send an I picture at a constant period. In this case, when the video content greatly changes in the P picture or the B picture, the data length of the P picture or the B picture at the change point increases.

FIG. 4 shows the G of the coded video data in this case.
This shows the OP structure, and the appearance period N of the I picture is 15 (sheets)
(A) shows the GOP structure when there is no scene change. FIG. (B) shows the case where the data length of the B picture at the position indicated by the arrow is increased due to the scene change. GOP structure.

In the second feature point detection method, a scene change point is detected by detecting an increase in the data length of a P picture or a B picture.

(First Method: FIGS. 5 to 7) FIG.
1 shows an example of the feature point detection unit 100 when a scene change point is detected by the method of (1).

As described above, the encoded video and audio data input from the external device 110 to the digital input / output terminal 19
/ TS, transmitted to the buffer control circuit 50 via the demultiplexing processing unit 40, and recorded on the disk 61 by the HDD 60. At the same time, the separation circuit 42 of the demultiplexing processing unit 40 converts the recorded MPEG / TS to the PES
Is extracted and supplied to the feature point detection unit 100.

The example of FIG. 5 can be applied to a case where the coded video data in the PES supplied to the feature point detection unit 100 has a GOP structure as shown in FIG.

The feature point detecting section 100 in this example includes a GOP structure analyzing section 101 and a video change point detecting section 102. The GOP structure analysis unit 101 detects the type (I, P, B picture) of each picture of the PES input to the feature point detection unit 100, and the video change point detection unit 102 From the picture type detected in, by the detection processing routine described below,
A scene change point is detected, and feature point (scene change point) information is output.

FIG. 6 shows an example of a detection processing routine performed by the video change point detection unit 102. In this detection processing routine, the processing is started with the count value n indicating the number of picture data from the immediately preceding I picture as shown in FIG. 7 being reset to 0. First, in step S101, the GOP structure analysis unit 101 The picture type is acquired, and then, in step S102, it is determined whether or not the picture at that time is an I picture.

If the picture at that time is an I picture, the flow advances to step S103 to determine whether or not the count value n is equal to the steady appearance cycle N of the I picture, ie, as shown in FIG. If N = 15, it is determined whether or not the count value n is 15.

As shown in FIG. 7, the count value n is set to 1 in the picture immediately after the I picture. Therefore, in the I picture, if there is no scene change, N (= 15)
, But does not reach N (= 15) if there is a scene change.

Therefore, when the count value n is N (= 15)
If the I picture is not a scene change point, the process proceeds from step S103 to step S104, where the count value n is set for the processing of the next picture.
Is set to 1, the process proceeds to step S107, the scene change detection flag is set to 0, and the process returns to step S101 to perform processing for the next picture.

If the count value n is not N (= 15),
Assuming that the I picture is a scene change point, the process proceeds from step S103 to step S105, where the count value n is set to 1 for processing of the next picture, and further proceeds to step S108, where the scene change detection flag is set to 1. After outputting the feature point information, step S
Returning to step 101, processing for the next picture is performed.

If it is determined in step S102 that the current picture is not an I-picture, the process proceeds to step S102.
Proceeding to 106, the count value n is incremented for processing on the next picture, and furthermore, step S1
In step 07, the scene change detection flag is set to 0, and the process returns to step S101 to perform processing for the next picture.

The feature point detection unit 100 in the example of FIG. 5 detects the scene change point from the PES as described above, and as the feature point information, indicates the time of the PES indicating the position of the I picture detected as the scene change point. PTS that is information
(PresentationTime Stamp) and a packet number are output. This feature point information is sent to the buffer control circuit 50 and recorded on the disk 61 while being managed separately from the MPEG / TS.

In the above example, the image change point detection unit 10
2 is configured by software, a dedicated hardware circuit may be used as the video change point detection unit 102.

(Second Method: FIGS. 8 to 12) FIG. 8 shows an example of the feature point detecting section 100 when a scene change point is detected by the second method.

The feature point detecting section 100 in this example includes a GOP structure analyzing section 101, a video change point detecting section 102, a threshold value determining section 103, and a memory 104. As the memory 104, a partial area of the RAM in the system controller 70 or the disk 61 is used instead of a dedicated memory.
May be used.

The GOP structure analysis unit 101 detects the type and data length of each picture of the PES input to the feature point detection unit 100, and the threshold value determination unit 103
The picture type and the picture data length are fetched from the P structure analysis unit 101, and the picture type and the picture data length of the m-th frame in the latest past are stored in the memory 104 by a threshold value determination processing routine described later.
Threshold L for data length of picture and B picture
Determine thP and LthB.

The video change point detection unit 102 calculates the picture type and the picture data length detected by the GOP structure analysis unit 101 and the threshold Lth determined by the threshold determination unit 103.
From P and LthB, a scene change point is detected by a detection processing routine described later, and feature point (scene change point) information is output.

The example of FIG. 8 can be applied to a case where the coded video data in the PES supplied to the feature point detection unit 100 has a GOP structure as shown in FIG. it can.

FIG. 9 shows a video change point detection unit 1 including a threshold value determination processing routine performed by the threshold value determination unit 103 in this case.
02 shows an example of a detection processing routine performed by the second routine. In this detection processing routine, the process is started with the count value a indicating the number of picture data stored in the memory 104 reset to 0. First, in step S111, the picture type and picture data The length L is obtained, and the thresholds LthP and Lth for the data lengths of the P-picture and the B-picture are described later by the threshold determination routine of step S120.
Determine B.

Next, in step S112, it is determined whether or not the current picture is a P picture. If the picture is a P picture, the flow advances to step S113 to determine the data length L of the P picture in step S120. Threshold Lth
It is determined whether it is greater than P, and if L ≦ LthP,
Assuming that the P picture is not a scene change point, the process proceeds to step S107, sets the scene change detection flag to 0, and returns to step S111 to perform processing for the next picture.

On the other hand, if L> LthP, it is determined that the P picture is a scene change point and step S11
From 3, the process proceeds to step S <b> 108, the scene change detection flag is set to 1 and the feature point information is output. Then, the process returns to step S <b> 111 to perform the process for the next picture.

If it is determined in step S112 that the picture at that time is not a P-picture,
Proceeding to 114, it is determined whether or not the current picture is a B picture.
Proceeding to S15, it is determined whether or not the data length L of the B picture is greater than the threshold LthB determined in step S120. If L ≦ LthB, the B picture is determined not to be a scene change point and step S107 is performed. After setting the scene change detection flag to 0, the process returns to step S111 to perform the processing for the next picture.

On the other hand, if L> LthB, it is determined that the B picture is a scene change point and step S11
From 5, the process proceeds to step S 108, sets the scene change detection flag to 1 and outputs feature point information, and then returns to step S 111 to perform the process for the next picture.

If it is determined in step S114 that the current picture is not a B picture, that is, it is an I picture, it is determined that the picture is not a scene change point, and the flow advances to step S107 to set the scene change detection flag to 0. Then, the process returns to step S111 to perform the process for the next picture.

The feature point detection unit 100 in the example of FIG. 8 detects the scene change point from the PES as described above, and indicates the position of the P picture or B picture detected as the scene change point as the feature point information. PES of PES
And packet numbers are output. As in the example of FIG. 5, this feature point information is sent to the buffer control circuit 50,
It is managed separately from the EG / TS and recorded on the disk 61.

Further, the example of FIG.
Is applied to both cases where the coded video data in the PES supplied to the PES has a GOP structure as shown in FIG. 3 and a GOP structure as shown in FIG. Can be.

FIG. 10 shows the threshold value determining unit 103 in this case.
4 shows an example of a detection processing routine performed by the video change point detection unit 102, including a threshold determination processing routine performed by the video processing apparatus. In this detection processing routine, the count value n indicating the number of picture data from the immediately preceding I picture and the count value a indicating the number of picture data stored in the memory 104 as shown in FIG. In step S111, the GOP structure analysis unit 101 sends a picture type and a picture data length L
Then, the thresholds LthP and LthB for the data lengths of the P picture and the B picture are determined by the threshold determination routine of step S120 as described later.

Next, in step S102, it is determined whether or not the picture at that time is an I picture. If the picture is an I picture, the flow advances to step S103 to set the count value n to the value shown in FIG. It is determined whether or not it is equal to the regular appearance cycle N of the I picture in the case of having the GOP structure as described above.
Assuming that the picture is not a scene change point, the process proceeds to step S104 to set the count value n to 1, and further proceeds to step S107 to set the scene change detection flag to 0.
Then, the process returns to step S111 to perform the process for the next picture.

If the count value n is not N, it is determined that the I picture is a scene change point, and the process proceeds to step S10.
From 3, the process proceeds to step S <b> 105 to set the count value n to 1, further proceeds to step S <b> 108, sets the scene change detection flag to 1, outputs feature point information, returns to step S <b> 111, and returns to step S <b> 111. Perform processing.

If it is determined in step S102 that the current picture is not an I-picture,
Proceeding to 106, the count value n is incremented, and then proceeding to step S112, it is determined whether or not the picture at that time is a P picture. Step S112
The following is the same as the detection processing routine of FIG.

In the case of FIG. 10, the feature point detection unit 1 of the example of FIG.
At 00, a PTS PTS and a packet number indicating the position of an I picture, P picture, or B picture detected as a scene change point are output as feature point information.

The threshold determination unit 103 in the example of FIG. 8 performs the threshold determination processing routine S120 in FIG. 9 or FIG. 10 based on the picture type and the picture data length of the latest m frames stored in the memory 104.
Threshold L for data length of picture and B picture
Determine thP and LthB. This is to more accurately detect an increase in the data length of the P picture or the B picture at the scene change point.

The value of m is the number of consecutive m's of encoded video data.
A value in which at least one P picture is included in a frame is assumed to be a value, and the coded video data is a GO as shown in FIG.
In the case of having a P structure, the number is 6 or more. FIG. 11 shows that m =
6 is assumed.

FIG. 12 shows an example of a threshold value determination processing routine S120 performed by the threshold value determination unit 103. In this threshold value determination processing routine S120, first, in step S121, it is determined whether or not the scene change detection flag set in step S108 or S107 in FIG. 9 or FIG. If the picture at that time is detected as a scene change point, and since there is no correlation between frames before and after the scene change point, the process proceeds from step S121 to step S122, where the contents of the memory 104 (picture After clearing the count value a indicating the number of picture data stored in the memory 104 to 0, and then proceeding to step S124, the scene change detection flag is cleared. If 0, directly from step S121,
Proceed to step S124.

In step S124, FIG. 9 or FIG.
In step S111, the picture type and the picture data length L acquired in step S111 are stored in the memory 104. Then, in step S125, it is determined whether or not the count value a has reached the specified value m, that is, as shown in FIG. If = 6, it is determined whether or not the count value a has reached 6.

Immediately after the detection processing of FIG. 9 or FIG. 10 is started or immediately after the count value a is reset to 0, the picture type and the picture data length L for m frames are still stored in the memory 104. Not a
If <m, the flow advances to step S126 to increment the count value a for the processing of the next picture, and then to step S127, where the threshold value LthP for the data length of the P picture and the data The threshold value LthB for the length is set to a sufficiently large value L such that the current P picture or B picture is not detected as a scene change point.
max, the threshold determination process ends, and the routine moves to step S112 in FIG. 9 or step S102 in FIG.

On the other hand, the picture type and picture data length L for m frames are stored in the memory 104 by the storage in step S124, and when the count value a reaches the specified value m, the process proceeds from step S125 to step S128. Then, after deleting the oldest picture type and the picture data length L from the memory 104 for the processing of the next picture, the process proceeds to step S129, and the threshold values LthP and LthB are calculated by the following equation (1).
The calculation is performed according to (2), and the threshold value determination processing is completed.
The process proceeds to step S112 of FIG.

LthP = Kp × LaveP (1) LthB = Kb × LaveB (2).

[0086] Each of SaveP and SaveB is stored in the memory 10 immediately before deleting the oldest picture type and picture data length L in step S128.
4 in the latest m frames in the past
The average value of the data length of the picture and the average value of the data length of the B picture, and Kp and Kb are one or more coefficients.

The coefficients Kp and Kb determine the scene change detection accuracy according to the values thereof. As the coefficient is closer to 1, even a slight video change is detected as a scene change. Is detected. Therefore, the coefficients Kp and Kb are set to 1 according to the type of video, the encoding algorithm of video data, and the like.
It is set to an appropriate value within the range of about 2 or less. Further, the coefficients Kp and Kb may be adjusted by the system controller 70 without being fixed values.

As described above, the average values of the data lengths of the P and B pictures in the latest m frames in the past P, L, P, L
The results of multiplying aveB by the coefficients Kp and Kb are used as thresholds LthP and LthB for the data lengths of the P and B pictures, so that the data of the P picture or B picture due to the scene change as shown in FIG. The increase in length can be detected more accurately.

The above example is based on the video change point detector 10 shown in FIG.
Although the second and threshold determination units 103 are configured by software, dedicated hardware circuits may be used as the video change point detection unit 102 and the threshold determination unit 103.

(Effect of Feature Point Detection of Embodiment) In the feature point detection method shown in FIG.
Is not detected from the baseband video data once decoded by the MPEG video / audio decoder 31 shown in FIG.
Since the scene change point is directly detected from the PES extracted from the EG / TS by the separation circuit 42, the amount of calculation for detecting the scene change point can be reduced, and the burden on the system controller 70 can be reduced. .

In the feature point detection method described above, a scene change point is detected without using the MPEG video / audio decoder 31, as described above. Can be used for applications.

That is, the coded video / audio data from the external device 110 is transmitted to the digital interface circuit 9.
0 to convert to MPEG / TS and record on disk 61,
From the MPEG / TS to be recorded, PES is performed by the separation circuit 42.
Is extracted by the feature point detection unit 100, and the feature point information is recorded on the disk 61. At the same time, the PES from the separation circuit 42 is supplied to the MPEG video / audio decoder 31 so that the output terminal 12V , 1
An analog video signal and an analog audio signal based on the video and audio data from the external device 110 can be obtained in the 2S and 12A.

Also, as the separation circuit 42, a circuit for detecting a scene change point from the PES by the feature point detection unit 100 and a PES from the MPEG / TS reproduced from the disk 61 and output to the MPEG video / audio decoder 31. When a supply circuit is provided, the encoded video / audio data from the external device 110 is converted into MPEG / TS by the digital interface circuit 90 and recorded on the disk 61. The change point is detected, and the feature point information is recorded on the disc 61, and at the same time, the MPE already recorded on the disc 61 is recorded.
G / TS is reproduced from the disk 61, and the reproduced MP
By extracting the PES from the EG / TS and supplying it to the MPEG video / audio decoder 31, the output terminals 12V, 1
MPE reproduced from disk 61 on 2S and 12A
A reproduced analog video signal and a reproduced analog audio signal by the G / TS can be obtained.

The feature point information recorded on the disk 61 is
Thereby, when playing back MPEG / TS from the disk 61, random access playback such as cue playback, playback from the beginning of a commercial message, playback from a specific scene in a program, and the like, can be realized. Thumbnails (reduced images) can be easily created.

[Other Embodiments] As shown in FIG. 1, the video signal from the terrestrial tuner 21 and the input terminals 11V, 1V
When the video signal from 1S is converted into digital video data by the NTSC decoder 25 and further encoded by the MPEG video encoder 27, and the MPEG / TS from the multiplexing circuit 41 shown in FIG. Also, the scene change point can be detected by the feature point detection unit 100 by the same method, and the feature point information can be recorded on the disk 61.

In the above-described embodiment, a hard disk is used as a recording medium (storage element) of the recording / reproducing apparatus. However, an optical disk, a magneto-optical disk, a semiconductor memory, or the like may be used.

In the above embodiment, the encoded video data and the feature point information are recorded on the same recording medium. For example, the encoded video data is recorded on a hard disk, and the feature point information is stored on a RAM. For example, the encoded video data and the feature point information may be recorded on different recording media.

[0098]

As described above, according to the present invention, a scene change point is directly detected from encoded video data by analyzing the picture structure of the encoded video data. Calculation amount can be reduced, and the load on the system controller can be reduced.

Further, since the scene change point can be detected without decoding the encoded video data, the video decoder can be used for another purpose when the scene change point is detected. The encoded video data recorded on the medium can be reproduced from the recording medium, decoded by a video decoder, and used to obtain a reproduced video signal.

[Brief description of the drawings]

FIG. 1 is a diagram showing an example of a recording / reproducing apparatus of the present invention.

FIG. 2 is a diagram illustrating an example of a demultiplexing processing unit.

FIG. 3 is a diagram illustrating an example of a GOP structure of MPEG encoded video data.

FIG. 4 is a diagram illustrating another example of a GOP structure of MPEG encoded video data.

FIG. 5 is a diagram illustrating an example of a feature point detection unit.

FIG. 6 is a diagram illustrating an example of a feature point detection processing routine.

FIG. 7 is a diagram provided for explanation of feature point detection in the examples of FIGS. 5 and 6;

FIG. 8 is a diagram illustrating another example of the feature point detection unit.

FIG. 9 is a diagram showing another example of a feature point detection processing routine.

FIG. 10 is a diagram showing still another example of the feature point detection processing routine.

FIG. 11 is a diagram provided for explanation of threshold value determination in feature point detection.

FIG. 12 is a diagram illustrating an example of a threshold value determination processing routine.

[Explanation of symbols]

Since the main parts are all described in the figure, they are omitted here.

──────────────────────────────────────────────────の Continued on the front page (51) Int.Cl. ⁷ Identification FI FI Theme Court ゛ (Reference) H04N 5/93 H04N 5/93 E (72) Inventor Tomoyuki Sato 6-7-35 Kita Shinagawa, Shinagawa-ku, Tokyo No. 7 Inside Sony Corporation (72) Inventor Makoto Mitsuno 6-7-35 Kita Shinagawa, Shinagawa-ku, Tokyo F-term within Sony Corporation (reference) 5C053 FA20 FA23 GB06 GB37 HA29 HA31 JA03 JA21 KA08 KA20 KA24 LA06 LA07 LA11 5C059 KK15 NN43 PP05 PP07 RC24 SS02 SS12 SS19 TA61 TB03 TC14 TC20 TD03 TD07 TD12 UA02 UA05 UA34

Claims (8)

[Claims] 1. A method for detecting a scene change point of an image as a feature point from video data encoded as a plurality of types of pictures, wherein an appearance cycle of a specific type of picture in the encoded video data is detected. A feature point detection method in which, when the appearance cycle is shorter than a predetermined cycle, a specific type of picture at that time is set as a scene change point. 2. A method for detecting a scene change point of an image as a feature point from video data encoded as a plurality of types of pictures, wherein each of the other types except a specific type in the encoded video data is provided. A feature point detecting method for detecting a data length of a picture of the other type and, when the data length becomes larger than a threshold value, setting another type of picture at that time as a scene change point. 3. The feature point detecting method according to claim 2, wherein the threshold value is a result of multiplying an average value of data lengths of pictures of the same type in a plurality of latest past frames in the encoded video data by a coefficient. Feature point detection method. 4. A feature point detection unit that detects a scene change point of an image as a feature point from video data encoded as a plurality of types of pictures, and outputs feature point information indicating a position of the scene change point. A coded video data and the feature point information, recorded on a recording medium, comprising a recording and reproduction unit for reproducing from the recording medium, the feature point detection unit, the appearance cycle of a specific type of picture in the coded video data A recording / reproducing apparatus that detects a specific type of picture at the time of detection and when its appearance cycle becomes shorter than a predetermined cycle as a scene change point. 5. The recording / reproducing apparatus according to claim 4, wherein said encoded video data is input from an external device to said recording / reproducing apparatus. 6. A feature point detection unit that detects a scene change point of an image as a feature point from video data encoded as a plurality of types of pictures, and outputs feature point information indicating a position of the scene change point, A recording / reproducing unit that records the encoded video data and the feature point information on a recording medium and reproduces the encoded video data from the recording medium, wherein the feature point detection unit is a type other than a specific type in the encoded video data. A recording / reproducing apparatus which detects a data length of each picture of the above and, when the data length becomes larger than a threshold value, sets another type of picture at that time as a scene change point. 7. The recording / reproducing apparatus according to claim 6, wherein the feature point detection unit multiplies an average value of data lengths of pictures of the same type in a plurality of latest past frames in the encoded video data by a coefficient. A recording / reproducing apparatus for calculating the threshold. 8. The recording / reproducing apparatus according to claim 6, wherein said encoded video data is input from an external device to said recording / reproducing apparatus.