JP2006254308A - Video recording apparatus and video recording method - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide a video recording apparatus and video recording method capable of automatically recording high-resolution video information. <P>SOLUTION: An inputted video signal is received and after only a moving region is extracted, only a part where the face of a human figure is photographed, is extracted from the extracted moving region. A frame where the face of the human figure is photographed best, is detected and a best-shot trigger signal is generated. The inputted video signal is frequency-converted into a frequency domain video signal, the frequency domain video signal is then divided into a plurality of sub band video signals and when the best shot trigger signal is generated, all the sub band video signals for a frame at that time are selected but when the best shot trigger signal is not generated, only sub band video signals in a part of a low frequency side corresponding to the frame at that time are selected. The selected sub band video signals are compressed as sub band compressed video signals, the sub band video signals are multiplexed therewith and recorded in a storage 112 as compressed video signals. <P>COPYRIGHT: (C)2006,JPO&NCIPI

Description

  The present invention relates to a video recording apparatus and a video recording method for compressing an input video and recording it on a recording medium such as a hard disk.

  2. Description of the Related Art Conventionally, there is a video recording apparatus that can switch resolutions by a user operation (for example, see Patent Document 1). In the moving image and still image recording / reproducing method disclosed in Patent Document 1, for example, a normal moving image frame is recorded with a pixel size of 320 × 240, but when a high image quality mode button is pressed, a frame image acquired at that timing is recorded. Can be recorded with a pixel size of 1280 × 960.

JP 2003-125344 A (6th page, FIG. 1)

  However, the conventional video recording apparatus has a problem that high-resolution video information cannot be automatically recorded because the resolution is switched by a user operation.

  The present invention has been made in view of such circumstances, and an object thereof is to provide a video recording apparatus and a video recording method capable of automatically recording high-resolution video information.

The above object can be achieved by the following constitution and method.
(1) A video recording apparatus for accumulating an input video signal in a recording medium, obtained by the frequency conversion means for converting the frequency of the input video signal to obtain a frequency domain video signal, and the frequency conversion means Subband dividing means for dividing the frequency domain video signal into a plurality of subband video signals, and for a video designated for high resolution recording, all of the plurality of subband video signals obtained by the subband dividing means are Selecting means for selecting a part on the low frequency side of the plurality of subband video signals obtained by the subband dividing means for the video that is selected and not designated for high resolution recording; and the selecting means Compression means for compressing each sub-band video signal selected in step (b) to generate a sub-band compressed video signal; and for the video designated for high-resolution recording, All the sub-band compressed video signals compressed by the compression means are multiplexed, and for the video that is not designated for high-resolution recording, the low-frequency side selected by the selection means and compressed by the compression means Multiplexing means for multiplexing some of the subband compressed video signals, and recording control means for recording the subband compressed video signals multiplexed by the multiplexing means on the recording medium. With this configuration, it is possible to automatically record high-resolution video information. In addition, since all of the video is not recorded at a high resolution and a video having no meaning is recorded at a low resolution, the recording medium can be effectively used.

(2) A video recording apparatus for storing a video signal input in units of frames in a recording medium, and performing image processing on the input video signal to detect a frame in which a person is best reflected to detect a best shot trigger signal A best shot trigger signal generating means for generating a frequency conversion means for frequency-converting an input video signal to obtain a frequency domain video signal, and a frequency domain video signal obtained by the frequency conversion means for a plurality of subband videos When the best shot trigger signal is generated by the subband dividing means for dividing the signal and the best shot trigger signal generating means, all the subband video signals corresponding to the frame at that time are selected, and the best shot trigger is selected. When no signal was generated, some subband images on the low frequency side corresponding to the frame at that time Selection means for selecting a signal, compression means for compressing each subband video signal selected by the selection means and outputting it as a subband compressed video signal, and each subband video signal compressed by the compression means And a recording control means for controlling the recording of the compressed video signal from the multiplexing means onto the recording medium. With this configuration, it is possible to record only an image in which a person's face is best reflected with high resolution. That is, it is possible to automatically acquire and record high-resolution video information from the input video signal. In addition, since all of the video is not recorded at a high resolution and a video that is not the best shot is recorded at a low resolution, the recording medium can be used effectively.

(3) A video recording method for storing an input video signal in a recording medium, wherein the input video signal is frequency-converted, and the resulting frequency domain video signal is divided into a plurality of subband video signals. For a video designated for high resolution recording, all of the plurality of subband video signals corresponding to the video are compressed to generate a subband compressed video signal, and all the subband compressed video signals obtained thereby are Multiplexed and recorded on a recording medium, and for the video that is not designated for high-resolution recording, a part of the low frequency side of the plurality of subband video signals for the video is compressed to generate a subband compressed video signal Then, some of the sub-band compressed video signals obtained thereby are multiplexed and recorded on the recording medium. With this configuration, it is possible to automatically record high-resolution video information. In addition, since all of the video is not recorded at a high resolution and a video having no meaning is recorded at a low resolution, the recording medium can be effectively used.

(4) A video recording method for receiving a video signal as input and storing it in a recording medium. After receiving an input video signal and extracting only a region with motion, a person is captured from the extracted motion region. The frame in which the person is best captured is detected to generate a best shot trigger signal, and the input video signal is frequency-converted into a frequency domain video signal. When the best shot trigger signal is generated by dividing into sub-band video signals, all sub-band video signals corresponding to the frame at that time are selected, and when the best shot trigger signal is not generated, Select a sub-band video signal on the low-frequency side corresponding to the frame of, and compress the selected sub-band video signal to sub-band compressed video No. As is recorded on the recording medium each sub-band video signals as multiplexed with the compressed video signal. With this configuration, an image in which a person's face is best reflected can be recorded with high resolution. That is, also in the present invention, high-resolution video information can be automatically acquired from an input video signal and recorded. In addition, since all of the video is not recorded at a high resolution and a video that is not the best shot is recorded at a low resolution, the recording medium can be used effectively.

  The present invention provides a video recording apparatus and a video recording method having an effect of automatically acquiring and recording high resolution video information from an input video signal by designating a video to be recorded at a high resolution. It is something that can be done.

  DESCRIPTION OF EXEMPLARY EMBODIMENTS Hereinafter, preferred embodiments for carrying out the invention will be described in detail with reference to the drawings.

  FIG. 1 is a block diagram showing a schematic configuration of a video recording apparatus according to an embodiment of the present invention. In this figure, a video recording apparatus 100 according to the present embodiment is used for monitoring to record a person, and includes a moving area extracting unit 101, a human area extracting unit 102, a best shot determining unit 103, A band division unit 104, a high frequency buffer 105, a low frequency buffer 106, a selection unit 107, a high frequency compression unit 108, a low frequency compression unit 109, a subband multiplexing unit 110, a recording control unit 111, And a storage 112.

  The subband division unit 104 corresponds to a frequency conversion unit and a subband division unit. The selection unit 107 corresponds to selection means. The high frequency compression unit 108 and the low frequency compression unit 109 correspond to compression means. The subband multiplexing unit 110 corresponds to multiplexing means. The recording control unit 111 corresponds to a recording control unit. In addition, the moving region extraction unit 101, the human region extraction unit 102, and the best shot determination unit 103 constitute a best shot trigger signal generation unit.

  The motion area extraction unit 101 receives a video signal from a camera (not shown) or a signal obtained by processing the video signal as an input, detects a motion in the video, extracts only the motion area, and extracts a motion area signal as a motion area signal. Input to the region extraction unit 102. As a method for extracting a moving region, the most common background difference method is used, but a method such as inter-frame difference, optical flow, or stereo vision may be used. As an output of the moving region extraction unit 101, a binary image is assumed in which a pixel having motion is “1” and a pixel having no motion is “0”.

  The human region extraction unit 102 determines whether the moving region is a person from the moving region signal input from the moving region extraction unit 101, and outputs the human region signal only when it is determined that the moving region is a person. That is, the human region extraction unit 102 operates only when there is movement in the video. If there is motion in the video, the motion region extraction unit 101 outputs a binary image that distinguishes between the region that has moved and the region that has not moved. It shall be determined whether or not. In order to apply the elliptical Hough method, edge detection is performed on the binary image input from the moving region extraction unit 101, and it is determined using the elliptical Hough transform whether the detected edge shape is close to an ellipse. . The elliptical Hough method is an effective technique that can accurately determine whether a person is a person because the head has an elliptical shape when a human body is captured in the video. As a method other than the elliptical Hough method, there is a method for detecting the structure of the face such as the eyes and nose.

  For example, a technique called boosting, which is often used in recent years, can detect a shadow of a face such as the eyes and nose at high speed by comparing with a very rough template. The advantage of the elliptical Hough method over the method of detecting the face structure is that a person can be detected with high accuracy even when the face structure cannot be clearly detected such as backwards. Depending on the camera installation position, the face of the passing person does not necessarily enter the field of view of the camera, so this time the elliptical Hough method is adopted. A method of detecting the face structure may be used as long as the face structure can be reliably photographed, such as a case where a camera is installed on the door.

  The best shot determination unit 103 determines the image quality of the person from the human area signal and instructs the best scene. That is, the best shot determination unit 103 selects one image in which a person is best captured from a series of videos. In the present embodiment, the best shot is selected based on the following four criteria: “person face size”, “face contrast”, “no missing”, and “face orientation”. The size of the human face is calculated from the matching template size in the elliptical hough. Further, if the fit with the elliptical template is partially biased, it is determined that there is a lack. For the orientation of the face, a luminance gradient in the lateral direction of the face is used.

  In general, the luminance gradient is flat in the front direction, and the luminance gradient occurs because the back of the head is dark and the face is bright in the right direction and the left direction. The face contrast can be calculated from the luminance width of the face portion. In the present embodiment, one best shot is selected using the majority of these four criteria. If only one criterion gets a high score, but many (three in this example) are better than the best shot parameters so far, the most recent frame is taken as the best shot and the previous best shot is replaced. To do.

  The subband division unit 104 converts the frequency of the input video signal into a frequency domain video signal, and divides the frequency domain video signal into a plurality of subband areas. That is, the input video is divided into certain subbands based on the frequency. In this embodiment, a horizontal two-stage division and a vertical two-stage division using a wavelet transform are assumed. FIG. 2A shows the result of dividing into two horizontal stages and two vertical stages. In this case, “H” represents the high frequency side, “L” represents the low frequency side, “LL” is a subband of the low frequency in the horizontal direction and the vertical direction, and “HL” is the high frequency in the horizontal direction and the vertical direction. A low-frequency subband, “LH”, is a low frequency band in the horizontal direction and a high frequency subband, and “HH” is a high frequency subband in the horizontal direction and the vertical direction. Since “LL” is a reduced image of 1/4 size with respect to the original video, it is an advantage of subband encoding that a reduced image used for thumbnail display or the like can be easily obtained. Further, by performing division in one horizontal stage and one vertical stage, it can be divided into seven bands as shown in FIG. The LLLL subband is a reduced image of 1/16 size.

  Each subband divided as described above is temporarily stored in a buffer. This is because it is necessary to wait until a person leaves the field of view to determine the best shot. The time until leaving depends on the scene used, but since it cannot be generally determined within a few seconds and the buffer capacity is also finite, processing when the maximum stay time specified by the buffer capacity is exceeded is required. Become. In this embodiment, when the maximum stay time is exceeded, the best shot up to that time is output as a confirmed best shot, and the best shot determination is restarted from the next frame. In the present embodiment, the LLLL subband shown in FIG. 2B is stored in the low frequency buffer 106 and the other six subbands are stored in the high frequency buffer 105.

  The selection unit 107 reads data from the high frequency buffer 105 or the low frequency buffer 106 according to the output of the best shot determination unit 103. That is, data is read from both the high frequency buffer 105 and the low frequency buffer 106 for the frame determined to be the best shot, and the low frequency buffer 106 is determined for the frame determined not to be the best shot. Read data from only. In the present embodiment, all the subbands (LLLL to HH) are output for the frame determined to be the best shot by the best shot determination unit 103, and the low frequency side for the frame determined not to be the best shot A part (LLLL) of subbands is selected and output.

  The best shot determination unit 103 outputs the number of frames determined as the best shot among the N frames stored in the buffer up to the present time. For example, when it is determined that the 22nd frame is the best shot while 30 frames are stored in the high frequency buffer 105 and the low frequency buffer 106, a numerical value “21” is output. Upon receiving this output, the selection unit 107 operates as follows. That is, since the 1st to 21st frames are not the best shots, the subbands of each of the 1st to 21st frames stored in the low frequency buffer 106 are read out and input to the low frequency compression unit 109. Further, the data of 1 to 21 frames stored in the high frequency buffer 105 is discarded.

  Since the next 22 frames are the best shots, 22 frames of subbands are read from both the low-frequency buffer 106 and the high-frequency buffer 105 and input to the high-frequency compression unit 108 and the low-frequency compression unit 109. In this case, the LLLL subband stored in the low-frequency buffer 106 is input to the low-frequency compression unit 109, and each of the LLHL, LLLH, LLHH, HL, LH, and HH stored in the high-frequency buffer 105 is stored. The band is input to the high frequency compression unit 108.

  Also, since the 23 to 30 frames are not the best shot as the 1 to 21 frames, the sub-bands from the 23 frames to the 30 frames stored in the low frequency buffer 106 are similarly read out. Input to the compression unit 109.

  The high frequency compression unit 108 compresses each subband input from the selection unit 107. As a result, the high frequency compression unit 108 operates only on the frame determined to be the best shot, and performs compression on the six high frequency bands. Then, the 6 subbands on the high frequency side subjected to compression are input to the subband multiplexing unit 110. As a result, the low-frequency compression unit 109 operates on all frames, and compresses the reduced image on the low frequency side. Then, the compressed reduced low-frequency image is input to the subband multiplexing unit 110. There are various compression methods, which are not particularly defined in the present invention, but in the present embodiment, compression by Huffman coding that determines a code length according to the appearance frequency of the code is assumed.

  The subband multiplexing unit 110 adds additional information such as a header to the data input from the high frequency compression unit 108 or the low frequency compression unit 109 and inputs the data to the recording control unit 111. For the frame determined to be the best shot, all seven subbands are processed, and for the other frames, only the reduced image on the low frequency side is processed. The recording control unit 111 records the data input from the subband multiplexing unit 110 in the storage 112. In particular, the type of storage is not specified, but in this example, a hard disk drive is assumed.

  Next, the operation of the video recording apparatus having the above configuration will be described with reference to FIG. In FIG. 3, the human head in the camera field of view is represented by a circle, and for ease of explanation, only the face size is used as the best shot criterion. For ease of explanation, a case of one horizontal stage and one vertical stage (for example, the case shown in FIG. 2A) is used.

  At the time of 0 frame input, the person has not yet appeared in the field of view. Since there is no person, the best shot determination unit 103 outputs “−1 (no person)”. As a result, in the 0th frame, only the low-frequency LL subband is compressed and recorded in the storage 112.

  When entering the first frame, a person appears in the lower left. The first frame is divided into subbands, the low frequency LL subband is held in the low frequency buffer 106, and the high frequency LH, HL, and HH subbands are held in the high frequency buffer 105. In this case, since there is no previous data, the reference of the first frame is set as a temporary best shot. The best shot determination unit 103 outputs “0”.

  When inputting the second frame, the face of the person becomes larger. The temporary best shot is updated to the second frame. The data of the second frame is divided into subbands, the low frequency LL subband is held in the low frequency buffer 106, and the high frequency LH, HL, and HH subbands are held in the high frequency buffer 105. The best shot determination unit 103 outputs “1”. Since the input of the second frame determines that the first frame is not the best shot, only the LL subband of the first frame is read from the low frequency buffer 106 and recorded in the storage 112 after being compressed. Is done. The high band side data of the first frame, that is, the LH, HL, and HH subbands are discarded.

  When inputting the third frame, the face of the person becomes larger. The temporary best shot is updated to the third frame. The data of the third frame is divided into subbands, the low frequency LL subband is held in the low frequency buffer 106, and the high frequency LH, HL, and HH subbands are held in the high frequency buffer 105. The best shot determination unit 103 outputs “2”. Since the input of the third frame determines that the second frame is not the best shot, only the LL subband of the second frame is read from the low frequency buffer 106 and recorded in the storage 112 after being compressed. Is done. The high band side data of the second frame, that is, the LH, HL, and HH subbands are discarded.

  When inputting the fourth frame, the face of the person becomes even larger. The temporary best shot is updated to the fourth frame. The data of the fourth frame is divided into subbands, the low frequency LL subband is held in the low frequency buffer 106, and the high frequency LH, HL, and HH subbands are held in the high frequency buffer 105. The best shot determination unit 103 outputs “3”. Since the input of the fourth frame determines that the third frame is not the best shot, only the LL subband of the third frame is read from the low frequency buffer 106 and recorded in the storage 112 after being compressed. Is done. The high band side data of the third frame, that is, the LH, HL, and HH subbands are discarded.

  At the time of inputting the fifth frame, the face of the person becomes larger, but part of the face goes out of the screen, resulting in omission. Therefore, the temporary best shot remains the fourth frame. The data of the fifth frame is divided into subbands, the low frequency LL subband is held in the low frequency buffer 106, and the high frequency LH, HL, and HH subbands are held in the high frequency buffer 105. The best shot determination unit 103 outputs “3”.

  At the time of inputting the sixth frame, the face of the person becomes smaller and missing is still occurring. Therefore, the temporary best shot remains the fourth frame. The data of the sixth frame is divided into subbands, the low frequency LL subband is held in the low frequency buffer 106, and the high frequency LH, HL, and HH subbands are held in the high frequency buffer 105. The best shot determination unit 103 outputs “3”.

  At the time of inputting the seventh frame, the omission is eliminated, but the human face becomes smaller. Therefore, the temporary best shot remains the fourth frame. The data of the seventh frame is divided into subbands, the low frequency LL subband is held in the low frequency buffer 106, and the high frequency LH, HL, and HH subbands are held in the high frequency buffer 105. The best shot determination unit 103 outputs “3”.

  When inputting the eighth frame, the person moves out of the screen. Therefore, the best shot is determined in the fourth frame. The best shot determination unit 103 outputs “−1”. The data of the eighth frame is divided into subbands, the low frequency LL subband is held in the low frequency buffer 106, and the high frequency LH, HL, and HH subbands are held in the high frequency buffer 105. .

  If the maximum stay time is exceeded at the input of the eighth frame, the best shot up to this point is output as the determined best shot. That is, the data of the fourth frame is the best shot, and all the subbands of the frame are compressed and recorded in the storage 112.

  In addition, the data of the fifth to eighth frames are recorded in the storage 112 after only the low-frequency subband is read and compressed. Further, the high frequency side data of the fifth to eighth frames is discarded.

  As described above, according to the video recording apparatus 100 of the present embodiment, after receiving an input video signal and extracting only a region where there is a motion, only a portion where a person's face is reflected is extracted from the extracted motion region. Extracting and detecting the frame in which the person's face is best captured, generating a best shot trigger signal, frequency-converting the input video signal into a frequency domain video signal, When the best shot trigger signal is generated after being divided into band video signals, all subband video signals corresponding to the frame at that time are selected, and when the best shot trigger signal is not generated, the frame at that time is selected. Select only a part of the corresponding sub-band video signal on the low-frequency side, compress each selected sub-band video signal, Since recording a subband image signal to the storage 112 as a multiplexed and compressed video signals, it is possible to record only an image in which the face of a person is reflected best in high resolution. That is, it is possible to automatically acquire and record high-resolution video information from the input video signal. In addition, since all of the video is not recorded at high resolution and video that is not the best shot is recorded at low resolution, the storage 112 can be used effectively.

  In this embodiment, for monitoring to record a person, a frame in which the person's face is best captured is recorded at a high resolution, and other frames are recorded at a low resolution. Is not limited to a person but is arbitrary. In short, it is only necessary to designate a video to be recorded at a high resolution, the designated video is recorded at a high resolution, and the others are recorded at a low resolution.

  The present invention has an effect of automatically acquiring and recording high-resolution video information from an input video signal, and compressing the input video signal and recording it on a recording medium such as a hard disk. Useful as such.

1 is a block diagram showing a schematic configuration of a video recording apparatus according to an embodiment of the present invention. Conceptual diagram for explaining the operation of subband division of the video recording apparatus of FIG. Conceptual diagram for explaining the operation of the video recording apparatus of FIG.

Explanation of symbols

DESCRIPTION OF SYMBOLS 100 Video recording device 101 Moving area extraction part 102 Human area extraction part 103 Best shot determination part 104 Subband division part 105 High frequency buffer 106 Low frequency buffer 107 Selection part 108 High frequency compression part 109 Low frequency compression part 110 Subband Multiplexer 111 Recording controller 112 Storage

Claims (4)

A video recording apparatus for storing an input video signal in a recording medium,
A frequency converting means for frequency-converting the input video signal to obtain a frequency domain video signal;
Subband dividing means for dividing the frequency domain video signal obtained by the frequency converting means into a plurality of subband video signals;
For the video designated for high resolution recording, all of the plurality of subband video signals obtained by the subband dividing means are selected, and for the video not designated for high resolution recording, the subband is selected. A selecting means for selecting a part on the low frequency side from the plurality of subband video signals obtained by the dividing means;
Compression means for compressing each subband video signal selected by the selection means to generate a subband compressed video signal;
For the video designated for high resolution recording, all the sub-band compressed video signals selected by the selection means and compressed by the compression means are multiplexed, and for the video not designated for high resolution recording. Is a multiplexing means for multiplexing a part of the sub-band compressed video signal on the low frequency side selected by the selection means and compressed by the compression means;
Recording control means for recording the sub-band compressed video signal multiplexed by the multiplexing means on the recording medium;
A video recording apparatus comprising: A video recording apparatus for storing a video signal input in frame units in a recording medium,
Best shot trigger signal generating means for generating a best shot trigger signal by detecting a frame in which a person is best captured by performing image processing on an input video signal;
A frequency converting means for frequency-converting the input video signal to obtain a frequency domain video signal;
Subband dividing means for dividing the frequency domain video signal obtained by the frequency converting means into a plurality of subband video signals;
When the best shot trigger signal is generated by the best shot trigger signal generation means, all subband video signals corresponding to the frame at that time are selected, and when the best shot trigger signal is not generated, then Selecting means for selecting a part of the subband video signal on the low frequency side corresponding to the frame of
Compression means for compressing each subband video signal selected by the selection means and outputting it as a subband compressed video signal;
Multiplexing means for multiplexing each subband video signal compressed by the compression means and outputting as a compressed video signal;
Recording control means for controlling recording of the compressed video signal from the multiplexing means onto the recording medium;
A video recording apparatus comprising: A video recording method for storing an input video signal in a recording medium,
The input video signal is frequency converted, and the resulting frequency domain video signal is divided into a plurality of subband video signals,
For a video designated for high resolution recording, all of the subband video signals for the video are compressed to generate a subband compressed video signal, and all the subband compressed video signals obtained thereby are multiplexed. For a video that is recorded on a recording medium and not designated for high-resolution recording, a part of the low frequency side of the plurality of subband video signals for the video is compressed to generate a subband compressed video signal. A video recording method for multiplexing a part of the subband compressed video signals obtained thereby and recording the multiplexed video signals on a recording medium. A video recording method for receiving a video signal as input and storing it in a recording medium,
After extracting only the region with motion in response to the input video signal, the best shot trigger signal is detected by extracting the part where the person is reflected from the extracted motion area and detecting the frame where the person is best reflected Produces
After frequency conversion of the input video signal into a frequency domain video signal, the frequency domain video signal is divided into a plurality of subband video signals,
When the best shot trigger signal is generated, all subband video signals corresponding to the current frame are selected. When the best shot trigger signal is not generated, the low frequency side corresponding to the current frame is selected. Select some subband video signals,
A video recording method of compressing each selected subband video signal as a subband compressed video signal and multiplexing each subband video signal and recording it as a compressed video signal on the recording medium.

Images