JPH0927944A - Moving image reproducing method - Google Patents

Abstract

PROBLEM TO BE SOLVED: To attain double speed reproduction for a moving image smoothly without decreasing number of totally delivered images. SOLUTION: When fast feeding reproduction or rewinding reproduction of any moving image information is requested from a terminal equipment A, referencing image frames Di-1, D1-2, D1-3, D1-4 belonging to an image group different from the moving image information are sequentially read of a storage device storing the moving image information and delivered to the terminal equipment. The delivered reference image frames are reproduced for a time interval smaller than a reproduction time interval of each image group in the case of usual reproduction. Thus, smoother reproduction than the conventional reproduction is realized. In this case, a ratio of a reproduction time interval to a 1 GOP reproduction time is limited to a range not in excess of a ratio of an I frame read time to a 1 GOP read time and the monopoly time of the storage device is not increased more than that at the usual reproduction.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a system for delivering moving image information from a server including a storage device for storing moving image information to a client connected by a communication line. The present invention relates to a moving image information reproducing method in which the number of simultaneous deliveries does not decrease and smooth reproduction is possible even when a terminal performing normal speed reproduction shifts to special reproduction when delivering and displaying to a terminal.

[0002]

2. Description of the Related Art In a moving image delivery system that delivers digitally compressed moving image information to a user terminal and expands and displays it, it seems that conventional analog video tapes are used to improve service to users and improve usability. Supporting special playback such as fast forward playback and rewind playback is an important function for performing content search easily and promptly. In the case of digital moving image information, the average bit rate for display on the terminal is specified by each compression standard, and using a higher display bit rate different from this at the time of special playback is a separate high-speed correspondence. Since a compression / expansion device is required, it is not preferable in terms of cost. Therefore, it is desired to realize special reproduction while keeping the display bit rate constant.

The following description and the embodiments of the present invention are based on the image format based on the MPEG standard. The MPEG is a prediction that is also compressed by using an image frame called an I frame, which is compressed only by spatial compression of the screen, and difference information from image frame information called P frames and B frames that are temporally preceding and following. One image group (GOP (group of pictures)) is composed of the frame and the frame. Hereinafter, the I frame will be referred to as a reference image frame, the P frame, and the B frame will be appropriately referred to as a prediction frame or an interpolation frame.

(Prior Art 1) The first method for realizing pseudo fast-forward reproduction / fast-reverse reproduction for performing a search while looking at the screen using MPEG is one image in the case of normal reproduction. Only the reference image frame (I frame) is reproduced and displayed for the time corresponding to the time when all the image frames in the group (GOP) are reproduced and displayed on the terminal, and the reference image frame number is skipped according to the requested double speed number. There is a method of realizing special playback such as fast-forward playback and rewind playback by going.

Regarding this method, the Industrial Development Organization Co., Ltd.
Information video of monthly magazine Page 19 or 2 of February 1992 issue
On page 7, there is a description that high-speed reproduction can be achieved by decoding I-frames and skipping I-frames. Further, it is described that the GOP length can be set freely on the encoding side. According to this method, it is considered that special playback is realized by the following method. The MPEG data structure will be separately described in detail in the embodiment with reference to FIG.

[0006] Now, one I frame, 14 P frames and B frames, and one G frame from a total of 15 image frames.
It is assumed that the OP is configured. The playback display is 1 /
It is assumed that the normal speed reproduction is realized by displaying one image for 30 seconds and successively performing the operations one after another. When performing double speed reproduction, the I frame in the first GOP is displayed for the reproduction time of 15 frames, and then the third G
This can be realized by displaying the I frame in the OP only for the reproduction time of 15 frames and reading and displaying the I frames in the fifth and seventh GOPs. In this case, a still image is displayed for 1/2 second which is a reproduction time of 15 frames. Further, by selecting the number of the I frame to be skipped, it is possible to cope with arbitrary speed.

(Prior Art 2) The second method is NT
T is “Technical Report of the Television Society” (March 1, 1994)
There is a method announced in "Highly Multiplexed Reading Method of Digital Video Information" posted on 8th. According to the method,
The image information is stored in stripes in order on the four magnetic disk drives for each divided information unit, and when the delivery request from the user, that is, the access to each disk drive is completely averaged. Is a system capable of channel distribution capable of normal speed reproduction for four disk drives. Although not explicitly shown in this example, it is assumed that one GOP or an integral multiple of GOP is assumed as a divided information unit.

In the above method, one GOP is used as a unit for reading and displaying, and skipping this GOP corresponds to fast-forward reproduction. The method disclosed in the above document is the following method. Special playback is realized by displaying an image after 4 seconds for 1 second during fast-forward playback and displaying a moving image for 1 second during reverse ascent 5 seconds before during fast-reverse playback. In this case, 5x speed fast forward reproduction and 5x speed rewinding reproduction are performed, respectively. In the literature, such a display method is referred to as skip reproduction.

The difference between this method and the first method is that a moving image is displayed for one second instead of a still image. However, according to the above-mentioned document, for example, when the image data of only one second is sequentially stored in each of the four drives, the selectable double speed number is n disks and skip segments p. , N × p + 1 times the speed.

In this prior art, the case of 5 × speed is shown. Image data is divided into GOP units and stored in order on four disks, and GOP numbers 1, 6, 11, and 1 are stored.
It is realized by reading out in the order of 6, ... When the GOPs to be read out in this order are sequentially read, the access shifts to the adjacent disks one after another, so the access becomes the same as the normal speed playback in which the access shifts to the adjacent disks in order. , It is possible to avoid the concentration of access to one certain disk.

[0011]

However, the above two methods have the following problems.

In the first method, since only one reference frame is reproduced at one GOP display time, smoothness is lacking in terms of connection before and after images.

The second method also has the same problems as the first method. Furthermore, since the achievable double speed number in the second method is determined depending on the number of drives, it is not possible to support arbitrary double speed. For example, 9x speed fast forward playback is 1
This is done by reading the tenth GOP after the th GOP. Here, if the fifth or sixth GOP which is temporally intermediate can be interpolated and displayed, the smoothness can be improved, but the second method does not have a function of interpolating and displaying by such an intermediate image frame.

Here, the possibility of interpolation display of an image will be touched upon. Generally, moving image information has a larger capacity than conventional text-based information, so it is considered advantageous to store it in a storage medium having a low bit cost and a relatively high transfer rate. As a medium used at present, a magnetic disk device is widely used because it can substantially satisfy both requirements.

In the magnetic disk device, the time required to position the magnetic head to the position where the information to be read is stored (generally called overhead time) takes several tens of milliseconds. Here, it is assumed that it takes 50 milliseconds. On the other hand, the amount of data to be read from the magnetic disk device at one time is one GOP, and this is read from a magnetic disk having a transfer rate of 5 MB per second, and the average bit rate required for display as an image compression method is about 200 kB. When MPEG1 of 1 is used, it takes about 50 milliseconds as a data read time. Therefore, the overhead time and the data read time are in the same order, and the read time determined by the sum of them is 100.
It takes about milliseconds. On the other hand, the reference image frame, without using the image frame information that is temporally before and after, only subjected to its own spatial compression, further because the reference image frame capacity is not particularly specified, Unless some limitation is imposed, the image content may reach almost half of the capacity of one image group. For this reason, even if only the reference image frame is read, it takes 75 milliseconds if the overhead time is included, and unless special measures are taken, during special playback, within a considerable time required to read one image group. Is limited to reading one reference image frame.

In order to eliminate the lack of smoothness, which is the drawback of the first or second method, the reading interval of the reference image frame must be made smaller than half of the reading interval at the normal reproduction. In order to do so, it takes more time to read a plurality of reference image frames than the time required to read each GOP during normal reproduction. As a result, the time available for image reading by other clients using the same moving image delivery system is reduced. As a result, the number of users who can receive the service of this system at the same time during double-speed reproduction is smaller than that during normal reproduction.

It is an object of the present invention to provide a moving picture reproducing method capable of smoother reproduction without increasing the time occupied by a storage device for holding moving picture information during double speed reproduction as compared with normal reproduction.

[0018]

In order to solve the above problems, according to the present invention, when a fast forward reproduction or a rewind reproduction of any moving image information is requested by a client, an image having different moving image information is displayed. A plurality of reference image frames belonging to the group are sequentially read from the storage device, and the plurality of read reference image frames are sequentially delivered to the client. In the client, these delivered plurality of reference image frames are reproduced at a time interval smaller than the reproduction time when each image group is normally reproduced during normal reproduction. As a result, at the time of fast-forward reproduction or rewinding reproduction, reference image frames having a larger number of frames than normal reproduction are reproduced, and smooth reproduction is realized. However, in the above time interval, the ratio of the time interval and the normal reproduction time of each image group is the average of the time required to read out the plurality of reference image frames belonging to the plurality of image groups of each moving image information (reference image Frame average read time) and the average time required to read each image group from the storage device (image group average read time)
It is set to be more than the ratio with. Due to this limitation,
The time occupied by the storage device does not increase as compared with the case of reading the moving image information for normal reproduction.

In the second invention of the present application, as the image frames forming each of the plurality of image groups of each moving image information,
An image group composed of a predetermined number of image frames larger than a predetermined reference number of image frames is stored. When a fast forward reproduction or a rewind reproduction of any moving image information is requested from the client, these images are skipped while skipping the interpolated image frames of the image groups having different moving image information, as in the first invention. A plurality of reference image frames of the group are sequentially read from the storage device, and the plurality of read reference image frames are sequentially delivered to the client. In the client, unlike the first invention of the present application, at a reproduction time interval smaller than the normal reproduction time when the client normally reproduces the image group including the image frames of the reference number of frames,
The delivered plurality of reference image frames are reproduced, and the reproduction time interval has a ratio between the time interval and the normal reproduction time of the image group of the reference frame number belongs to the plurality of image groups of each moving image information. The average time required to read each of the plurality of reference image frames (average read time of the reference image frames) and the average time required to read each image group of the moving image information from the storage device (image group The average read time) is determined to be equal to or greater than the ratio.

By increasing the number of frames belonging to each image group, the ratio of the time for reading each reference image frame to the time for reading the entire image group is reduced. However, since the number of frames in each image group is increased, it is possible that the reference image frame is simply reproduced at a time interval shorter than the reproduction time of each image group.
This reproduction is not always performed at a time interval shorter than the reproduction time of the image group having the conventional reference frame number. Therefore, in this case, the image group of the reference number of frames is reproduced at a time interval shorter than the reproduction time, and the lower limit of the reproduction time is set from the image frames of the reference number of frames to the time when the storage device is occupied. No larger than when playing a group of images.

More specifically, the features of the present invention will be described. According to the present invention, when special reproduction such as fast-forward reproduction or fast-reverse reproduction is performed, a plurality of sheets are recorded from the storage device in the delivery server during normal speed reproduction. The following means are used to read at least two screens of the reference image frame corresponding to the time required to read one GOP data composed of the image frame information.

The delivery server controls a storage device that stores image information, a storage control device that controls writing and reading of the storage device, a communication device that transmits information to a terminal, and the whole of these components. It is composed of a central processing unit, etc. The central processing unit includes software for selecting a GOP number or an I frame number to be read from the storage device in response to a title delivery request from the terminal or a double speed number.

One or more terminals are connected to one delivery server to request information from the user and display expanded image information. The components include a communication device for transmitting a request to the delivery server, receiving transmission information from the delivery server, a decompression device, a buffer for holding information to be displayed, a display device, and the like. The display time of the decompressed information and switching to the subsequent image are performed by software processing in the central processing unit in the terminal.

According to the present invention, with the above configuration, in the delivery server, the number of frames constituting the GOP and the I within the GOP are set.
Arrangement of frames, storage arrangement of GOP in disk, selection of number of GOP read during normal reproduction, selection of read I frame number during double speed reproduction, and fast forward reproduction by controlling display time of I frame at terminal. Achieve special playback. In the following description, the case where the number of frames forming one GOP is 15 will be used as the reference state.

The first implementation method will be described below. GOP
The number of constituent frames is unchanged from the standard 15
This is a method that is realized by devising the arrangement within the disc. Here, the I frame is placed at the head of the GOP. GO
The storage arrangement of P in the disk is such that GO is continuous in time.
Store P physically consecutively within the disk. During normal reproduction, two consecutive GOPs are continuously read. Since consecutive GOPs are physically consecutively arranged, the overhead time such as head movement time and rotation waiting time that occurs between two GOPs can be reduced to a negligible extent. Even when reading a plurality of I-frames of a predetermined number corresponding to the double speed number during special reproduction, the I-frame is arranged at the head of the GOP, so that the overhead time at the time of reading the second and subsequent I-frames can be shortened. During fast-forward playback at a very high-speed magnification, the I-frame number to be read will be large, which may cause an increase in overhead time. Generally, fast-forward playback while looking at the screen is at most about 10 times. However, if the speed is higher than this, it may be easier to use by jumping to an appropriate position on the title. On the other hand, the conventional method is to display and display more than one I frame on average for the reproduction display time of one GOP, that is, the reproduction display time of 15 frames in this case. Compared with the case of 1, smoothness can be improved. Detailed description will be given in the description of the embodiment.
In addition, the reproduction display interval is limited so that the time for occupying the storage device during the fast-forward reproduction does not substantially increase than that during the normal reproduction. In this method, the I-frames in two discontinuous GOPs are read at a speed of 3 times or more, and therefore the average time for reading the I-frames may be larger than the average time for reading the entire GOPs. Occurs. Therefore, the time for monopolizing the storage device increases as compared with the case of the double speed.

The second implementation method will be described below. GOP
This method is realized by defining the number of constituent frames and the capacity of I frames. It is specified at the time of compression so that the number of image frames forming the GOP is increased from the reference state and the data capacity of the I frame is reduced as compared with that in the reference state. As a result, the average time required to read each I frame (average read time of the reference image frame)
Is set to be smaller than that in the first method from the average of the read times of each GOP (the average read time of the image group). In order to improve the smoothness of reproduction during special reproduction in this case, the reproduction time interval of the reference image frame is set to be shorter than the reproduction time of the GOP composed of the reference number of frames. At this time, in order to prevent an increase in the exclusive time for the storage device, the reproduction time interval is set such that the ratio between the reproduction time interval and the normal reproduction time of the image group of the reference number of frames is equal to Of the average time required to read each of the plurality of reference image frames belonging to the image group (average read time of the reference image frames) and the time required to read each image group of the moving image information from the storage device. The value is set to be equal to or greater than the ratio to the average (average read time of image group). As a result, the time for monopolizing the storage device during special reproduction does not become longer than during normal reproduction of an image group consisting of image frames of the reference number of frames.

The fact that the number of delivery channels is the same during normal speed reproduction and special reproduction facilitates system design, and when a user shifts from normal speed reproduction to special reproduction and returns to normal reproduction again. However, if the total number of delivery channels does not change, delivery can be continued without problems such as delivery restrictions to other user terminals.

In any of the implementation methods, image information about the number of image frames constituting the GOP, the storage position, and the storage position of the I frame is recorded as a table on the disc when the image is stored on the disc, and is reproduced at the time of reproduction. By reading out these image information, the data arrangement can be recognized in advance, which enables high-speed disk access.

The moving image information is stored in a storage device which is a storage means by using a GOP composed of a plurality of frame information including at least one I frame as a storage unit, and in addition to the reading process for each storage unit, Information such as the storage position and the number of storage frames is added to the image information in advance and stored so that the reading can be performed for each I frame. When storing image information in a single disk device, successive image groups are first placed on adjacent disk surfaces, and after the same cylinder is full, subsequent information is placed on adjacent cylinders. When storing image information in a plurality of disk devices, the image groups are cyclically arranged in disk devices in units. High-speed access to an arbitrary GOP or I frame is realized based on the above storage arrangement information. Reading at normal speed and delivery to the user are performed for each image group, and during special playback,
Only I-frames are played sequentially. As a result, the number of users that can be delivered simultaneously is increased and the overhead time is shortened equivalently.

In the first recording method of the above-mentioned two recording methods, particularly in the case of double speed special reproduction, the average value of the time for reading each I frame of two adjacent GOPs is one GOP. It can be close to half the average value of the entire read time. However, at 3 × speed or more, since I frames in two discontinuous GOPs are read, the average value of the time for reading the I frames may become larger than the average value of the time for reading each GOP as a whole. However, the I frames of these GOPs can be selectively read out much faster than when two GOPs to be read successively are not recorded in the same cylinder. The second recording method described above has the same problem.

As described above, according to the present invention, in a device for storing and delivering a moving image, when delivering a plurality of titles simultaneously overlapping to a plurality of users, it is possible to access when performing continuous reproduction at normal speed. It enables smooth special playback for all users who are accessing without reducing the number of terminals.

[0032]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS The moving image information reproducing method according to the present invention will be described below in more detail with reference to some embodiments shown in the drawings. In the following, the same reference numerals represent the same or similar ones.

<Embodiment 1> (Apparatus configuration) FIG. 1 is a diagram showing the method of the present invention.
It shows a device configuration when information is delivered from the delivery server side to the terminal (client) side. Although only one terminal is shown in FIG. 1, even if there are a plurality of terminals 150, the basic configuration does not change except that a plurality of delivery channels of the communication control unit 130 to the terminals correspond.

In FIG. 1, examples of information handled as continuous information include moving images and sounds. The information input device 101 corresponds to a device such as a camera or a microphone that directly captures the occurrence of an event, or a medium such as a tape or a disk in which information is stored in advance. Then, the compression device 102 reduces the capacity of the original information as necessary. After this,
It is recorded in the storage device 103. When recording, management information related to moving image information is added. These additional information will be described later. The information recorded in the storage device 103 is delivered after being converted into an information format that can be transmitted to the communication line 140 by the communication device 104 in response to a delivery request from the terminal 150. Each of the above-mentioned devices constituting the delivery server 100 is controlled by a central processing unit 101 (generally a CPU; central processing unit). The input control device 12 corresponds to each of the devices 101 to 104.
1, the compression control device 122, the storage control device 123, and the communication control device 124 intervene to control each element. A memory 102 for executing a control program is connected to the central processing unit 101. In the present invention, the storage control software 10 which is the center of control is described as existing in the memory 102 in FIG. The control software for other components also exists in the memory 102 and is executed.

Here, the communication line 140 may be either a long-distance communication line such as a coaxial cable or an optical fiber, or a short-distance communication line such as a pair line or a shield line. Generally, the distance between the server and the terminal. Selected by. Further, the communication protocol may be a general-purpose protocol such as ATM (Asynchronous Transmission Mode) in which each tens of bytes is packetized and each packet is attached with a header indicating a delivery destination.

The communication device 1 is used as an information input unit for the terminal 150.
The information transferred by the communication control device 171 is received according to the communication protocol. Since the transmitted information is already compressed by the compression device 102,
The decompression device 152 for performing the operation opposite to the compression restores the original information form. Since a moving image is used as the information here, the image is displayed on the display device 154 via the buffer 153. Similar to the delivery server 100, the central processing unit 161 controls the entire terminal. The input device 155 has a function of inputting a command from a user.
Each of the devices 151-153 has a control device 171-
The operation is controlled by 173. Also in the terminal 150, as in the case of the delivery server 100, the decompression display control software 11 for controlling the decompression reproduction time and the like exists in the memory 162, and the time count for display by the control software,
Performs processing such as screen switching.

(Storage Method) FIG. 2 is a diagram for explaining the method of storing the image information used in this embodiment in the disk device. In the present embodiment, two or more reference image frames (I frames) are read within the time corresponding to the time of reading one image group (GOP). For this purpose, it is necessary to optimize the storage method to reduce the overhead time.
In short, continuous GOP data corresponds to physically continuous storage in the disk.

In FIG. 2, information storage in a single magnetic disk drive will be described as an example. In FIG. 2A,
Three titles, title 1 to title 3, are time-sequentially Da-1 to Da-n (where a is the title number,
It is assumed that n is divided up to the division section number of the title. For ease of handling, the unit of this division is GOP. Consider a state where these titles are continuously stored on the disc. For example, when MPEG1 having an average transfer rate required for display of 1.5 megabits per second (about 200 kilobytes per second) is used as an image compression method, and when a disk device having a capacity of 2 gigabytes is used, data of about 10,000 seconds is stored. Can accumulate. If one title is information for one hour, it is an account that can store about three titles.

A general magnetic disk drive is assumed as the disk device. In the magnetic disk drive shown in FIG. 2B, a plurality of magnetic recording disks are rotated by a common spindle shaft, and the information recorded on the recording disks is read by magnetic heads provided on the respective surfaces. . The reproduced signal from each magnetic head is selected by the switch and input to the subsequent signal processing circuit. There are multiple magnetic heads, but only one is active at the same time.

As an example, consider a case where the information D1-1 is recorded on the first surface and the information D1-2 following the information is recorded on the second surface of a plurality of recording disks. In FIG. 2C, assuming that the disk is rotating clockwise as viewed from the upper surface of the paper and the magnetic head is fixed in the circumferential direction, as shown by the arrow in FIG. The information up to b) is read. The magnetic head is switched when the reproduction of the information recorded on the second surface is started. If this switching operation is completed within the time required to move the angle from (b) to (c) in the figure, the information can be read from (a) to (d) without any break in information.

FIG. 2D is a diagram showing an example of correspondence between the physical format on the disc and the image information. Recording of information on the disc and reproduction of information from the disc are performed in units of sectors. Although omitted in the figure because it becomes complicated, header information including position information within the disc is attached to the beginning of the sector, and the actual data is recorded in the area after this. The sector shown in FIG. 2D shows only the data area excluding the header. When focusing on one surface, one round of the disk is called a track. One track is composed of a plurality of sectors. In the figure, the case where one track is composed of m sectors is shown. Since a magnetic disk is generally composed of a plurality of disks, a cylinder is formed by combining tracks on different surfaces. The cylinder is named because it consists of a plurality of tracks located at the same radial position and has a cylindrical shape. FIG.
In (d), an example is shown in which one cylinder is composed of N tracks. For example, if there are 5 disks and all of them can be used as data surfaces, there are 10 data surfaces, so N becomes 10.

To store moving image information in a disk device having such a physical format structure, the moving image information is sequentially stored in consecutive sectors as shown in FIG. Of different tracks, that is, different surfaces located at the same radius, are successively stored. Normally, the time required for the head to move between tracks requires about 2 milliseconds for one track, but by storing the data on another disk surface in consideration of the switching time as shown in FIG. 2C, The time required to move to another track can be made almost invisible.

FIG. 2E shows an example of storing consecutive GOPs. Basically, it corresponds to the expanded view of FIG. In FIG. 2E, the track is track 0.
8 tracks from 8 to 7, that is, FIG. 2 (d)
Then, N corresponds to 8, and cylinders exist from cylinder 0 to cylinder S. As shown in the figure, the GOP is first stored in consecutive sectors of the same track, and when one track is exceeded, storage is continued from the head sector of the following track. In the figure, the place described as a dummy is a GOP
This means that if the break in the middle of a sector is in the middle of a sector, the latter half of the same sector is filled with dummy data. Since writing and reading to and from the disk are performed in units of sectors, matching the beginning of the GOP with the beginning of the sector has the effect of facilitating the reading process. GOP storage after one cylinder is full moves to the next cylinder number. When the magnetic head crosses over the cylinders, the magnetic heads move to the cylinders, but this takes about 2 ms or less, which is 10% or less of the overhead time described later. Therefore, a large increase in the read processing time occurs. Absent.

(Reading Method) Next, a delivery method to the terminals in the configuration shown in FIG. 1 will be described. FIG. 3 is a diagram showing a reading method from a disc and a display operation on the terminal side when delivering to a plurality of terminals.
When the selection of moving image information is input from the terminal 150 by the input device 155, the central processing unit 161 interprets the request content, and the communication line 140 is transmitted from the communication device 151 via the communication control device 171 in a data format according to the communication protocol. Through the delivery server 100. The transmitted delivery request is received by the communication device 104, interpreted by the central processing unit 101 via the communication control device 124, and stored in the storage control device 12.
Target image information is read from the storage device 103 controlled by 3. In FIG. 3, a case where the magnetic disk described in FIG. 2 is used as the storage device 103 will be described as an example. FIG.
As will be described in more detail below, information such as the arrangement of each title and the storage position of GOP is recorded in advance on the disc as a table, and the table is read out from the disc onto the memory 102 at the time of system startup. When the reading of the following information is started, the storage position can be known each time without accessing the disk.

In FIG. 3, moving image information is read from the disc in the [server]. At this time, it is assumed that the same title is physically contiguously stored in the disc in units of GOP as already described with reference to FIG. In the example of FIG. 3, when performing normal speed reproduction,
It is assumed that two GOPs including the reference image frame are collectively read. In FIG. 3, terminal A and terminal C are performing normal speed reproduction, and terminal B is performing double speed reproduction. In the case of fast-forward reproduction with a relatively low magnification, the I frame to be read is arranged in the vicinity in the disc. Therefore, the overhead time immediately before reading the second I frame is significantly reduced.

Here, in order to quantify the explanation of the read processing time, it is assumed that the average transfer rate required for display is MPEG1 of 1.5 megabits per second. Although it will be mentioned again later in the description of the second embodiment of the present invention, FIG. 4 shows an explanatory diagram of the read processing time. FIG. 4B corresponds to a diagram showing the disc read time relationship of the first embodiment shown in FIG. When reading the information stored in the magnetic disk, there is an overhead for positioning the magnetic head in the sector in which the target image information is stored. The overhead is determined by the disk rotation waiting time, the head seek time, the time required for command transmission and processing via the disk interface, and the like. As a numerical example, the rotation speed is 540 rpm.
Assuming that the number of rotations is 0, the average rotation waiting time is 7 milliseconds as the time required for one half rotation, the average head seek time is 10 milliseconds, the command transmission processing time is 3 milliseconds, and there is a total overhead time of 20 milliseconds. I assume. The data transfer rate of the disk is 5 megabytes per second.
Consider a case where the image group is composed of 15 frames. If 30 frames per second are required to display an image, the capacity is about 100 kilobytes from the average transfer speed required for display, and the read time from the disk requires about 20 milliseconds. Also, the maximum capacity of an I frame is 1/2 of that of a GOP composed of 15 frames.
Is assumed to be.

FIG. 4A shows the read time relationship in the conventional method. In the conventional method, it is assumed that GOPs are stored without being aware of continuous storage. Applying the above values, the overhead (abbreviated as OH in the figure) time is 20 milliseconds to read one GOP during normal playback, and GO
The P read time is 20 milliseconds, which takes a total of 40 milliseconds. Also, during special playback such as fast-forward playback, the overhead is 20
It takes a maximum of 10 milliseconds to read an I frame in milliseconds. Even when the second I frame is read, the overhead time is still 20 milliseconds because it is stored without considering the continuity, and then the I frame is read in 10 milliseconds. After all, it takes 60 milliseconds in total. Therefore, during special reproduction, the processing time is extended by about 20 milliseconds, and the total number of delivery channels is also reduced to 2/3. On the other hand, in FIG. 4B which is the first embodiment of the present invention, two GOPs can be read in a state where the overhead time between GOPs can be almost ignored after the overhead time of 20 milliseconds during normal reproduction. It takes 60 milliseconds in total. Also, during special playback, one I-frame is read out in 10 ms after an overhead of 20 ms, and the overhead that occurs when the subsequent I-frame is read out is about 7 ms of rotation waiting time and seek for one track. Since the time is less than 2 milliseconds, the overhead here is less than 10 milliseconds,
After all, the processing is completed in about 50 milliseconds. For this reason, when the number of delivery channels of the system is designed with reference to the time of normal reproduction, special reproduction can be executed without reducing the number of channels.

The present invention improves the connection or smoothness of the screens during special reproduction, and provides the number of channels that can be specially reproduced without reducing the number of deliverable channels during normal speed reproduction. There are two points. This point will be quantified and described in the first embodiment of the present invention.

First, regarding the reference state, FIG.
This will be described with reference to FIG. The smoothness at the time of special playback is 1 in the standard state, that is, 15 frame display time.
It is considered that this can be improved by reproducing and displaying I frames more frequently than reproducing and displaying one I frame. At this time, I is used at an α times frequency of 15 frames in the reference state.
The reduction rate R of the number of delivery channels when reading a frame is given by the following equation.

R = (OH time + I frame read time) × (number of GOP frames / 15 frames) × α / (OH time + GOP read time) (Equation 1) In the case of the first embodiment of the present invention, Equation 1 is , R = (0.5 × OH time + I frame read time) × (15/15) × α / (0.5 × OH time + GOP read time) = (0.5 × 20 ms + 10 ms) × 1 × α / (0.5 × 20 msec + 20 msec) = 2α / 3 (Formula 2) Here, the condition that the number of delivery channels is not reduced during special reproduction is R ≦ 1. Therefore, substituting into Equation 2, α ≤
It becomes 3/2. This corresponds to reading the I frame at a frequency of 3/2 times higher than reading one I frame for every 15 frames in the reference state. That is, it can be seen that the number of delivery channels does not decrease during special reproduction until the condition of reading one I frame for 10 frame reproduction time is reached. In the first embodiment, as shown in FIG. 4, the average read time of the reference image frame is 25 ms, and the average read time of one GOP is 30 ms. Therefore, at the time of fast-forwarding, the I-frame playback interval is shortened by 30/25 times (1.2 times) the ratio of these numerical values, and even if the disk storage device is frequently accessed, the exclusive time of the disk storage device is reduced. , Will not increase more than normal playback. Therefore,
In the present embodiment, the lower limit of the reproduction time interval is set such that the ratio of this time interval and 1 GOP reproduction time is 1.2 or more.
After all, in the present embodiment, since the reproduction time interval is not half of the 1 GOP reproduction time, the reproduction is not exactly double speed but exactly 1.2 times speed. However, in the following, such reproduction may be simply referred to as double speed in the sense that it is close to double speed. Of course, if the disk access overhead is reduced, the double speed in the accurate sense can be realized.

FIG. 3 shows an example in which two GOPs are continuously read out. In this figure, α is shown only when it is 1. However, when a certain terminal is viewed as an average for a long time, it is higher in frequency than the reference state. The I frame can be played back and displayed.

The recording method used in this method is effective as it is when reading I frames in two consecutive GOPs as in the case of double-speed reproduction, but in the case of triple speed or more, it is effective in two discontinuous GOPs. Since the I frame is read, the average time for reading the I frame approaches the average time for reading the entire GOP. However, even in the case of 3 × speed or more, the average value of the time required to read the I frame is more than the time required to read the entire 1 GOP than the case where each GOP is stored in different storage locations. Can be small.

(Expanded Display Method) Returning to FIG. 3 again, expansion and display operation on the terminal side will be described. In FIG. 3, 2
2 GOPs for one GOP or two I-frame information
After reading at the average read time and delivering these pieces of information to the terminal, the information is expanded by the expansion device 152 provided in the terminal,
The data is temporarily stored in the buffer 153 and displayed on the display device 154. Since the data read speed of the disk is faster than the speed required for display on the terminal side, the number of deliverable channels (the number of terminals) is determined by the ratio of the read speed and the playback display speed.

In order for the delivered image to continue to be displayed without interruption, it is necessary to start delivery of the GOP (I frame during special playback) that continues before decompressing the compressed information delivered to the terminal. The 2 GOP average reproduction display time in FIG. 3 corresponds to the average time required for decompressing the delivered two GOP total data on the terminal side.

In FIG. 3, three titles D1, D2 and D3 are shown.
Are requested by terminal A, terminal B, and terminal C, respectively, and terminal B requests the title D2 for fast-forward reproduction at double speed. To the terminal A requesting the title D1, the title D1 is delivered from the server within each cycle time in the order of D1-1, D1-2, D1-3 in the segment order. Similarly, to terminal C requesting title D3, it is delivered in the order of D3-1 and D3-2. Here, for the terminal B requesting the double speed reproduction, two reference frames can be reproduced at a time corresponding to the time for expanding and displaying the image group (for example, D1-1) on the other terminal. That is, the reference image frame information in the image groups of D2-1 and D2-3 is read from the disk within one terminal allotted time, the reference frame information delivered to the terminal B is expanded, and the time corresponding to half the normal reproduction time is reached. It is held in the buffer 153 up to and displayed on the display device 154. After that, the reference image frame information in D2-3 that has been read in advance is decompressed and is also held in the buffer.
Continue to display. Since the time for displaying one frame in normal speed playback is regulated to be 1/30 second, the target I can be used even in special playback by counting the time on the terminal side.
The frames can be appropriately switched and displayed. In FIG. 3, decompression and display are performed in parallel in terminal B that is performing double-speed display, but this processing can be implemented by using buffer 153 as a two-sided buffer.

(Data Structure) Next, the data structure and data format of the moving image information will be described.

FIG. 10 shows M, which is one of the moving image compression methods.
It is the figure which showed the data structure regarding a PEG (Moving Picture Expert Group) system. The transfer rate required to display normal NTSC standard moving image information is about 200 Mbps (megabits per second). In this state, a huge amount of capacity is required for storage and transmission, so various compression methods are used. To be The MPEG system uses both temporal compression and spatial compression to reduce the transfer rate to an average of 1.5 Mbps in MPEG1 and an average of approximately 6 Mbps in MPEG2. As shown in FIG. 10, in the moving image data, an image group GOP including a plurality of frame images is recorded together with the sequence header SH. SH is the horizontal and vertical size for the following GOP,
Information such as aspect ratio and frame rate is described, and the operation of the MPEG decoder for decompression is determined based on the information. In the sequence header SH, there is user data UD that can be uniquely used by the vendor or the user after the original SH included in the MPEG standard. This UD section can also be used to describe control information when performing a unique function or operation not described in the MPEG standard. A specific method of using the UD portion in the present invention will be described later.

A GOP is generally composed of a plurality of image frames, and is defined by the MPEG standard as an image group including at least one reference frame (I frame). The I frame is only spatially compressed within the frame, and the image can be expanded without using information of other image frames existing before and after the I frame. The I frame is used to start image display when displaying an image. In MPEG, temporal compression is also used in order to increase the compression rate as described above. The prediction frame (P frame) and the bidirectional prediction frame (B frame) are temporally compressed. The P frame is frame information constructed from the information of the I frame and the information before the P frame in terms of time. The B frame is frame information constructed from the P frame information and the I frame information that are past and future in time with respect to the B frame. The data size after compression will increase or decrease depending on the detail of the image, etc.
In the case of MPEG1, encoding is performed so that the average transfer rate is 1.5 Mbps.

The ratio of the I frame information in one GOP varies depending on the number of frames in the GOP, the image content, etc., but there is no particular limitation, and generally 3 to 4
It is said that it is relatively cheap. Each image frame is further sliced into a plurality of data blocks and divided into a plurality of macroblocks MB.

(Related Information Table) In the apparatus configuration shown in FIG. 1, when the information is stored by the method of FIG. 2 and delivered by the method of FIG. 3, the position in the disk where the image information is stored in advance is determined. Need to figure out. FIG. 5 (a)
(B) is an implementation example of the information structure for indicating the image information storage position. FIG. 5A is a diagram showing the structure of the image information table. Required information includes stored title information 320 and image group (GOP) information 32.
1, and the picture in the GOP. This is stored information 322. The title information 320 is composed of a data type 301 indicating what the stored information is, such as a movie, news, video mail, etc., and a title number 302. GOP information 3
21 is each GOP number 303, data size 304,
The number of frames is 305. Picture. The storage information 322 includes the storage address 306 in the disc of the reference image frame (described as I picture in the figure), the size 307,
Further, it is composed of a data block size 308 read by one I / O request, and a disk block address (logical address of the storage position in the disk) 309. FIG. 5B is a diagram showing a state in which image information is stored for each block of the disc. In the example of FIG. 5B, each data block size is fixed. A block header 310 and a sequence header 311 are added to the beginning of each image block. A GOP header 312 is added to the beginning of the GOP, and a picture header 313 is added to the beginning of each image frame so that the start of the information can be recognized. The means for recording these data formats is included in the functions of the conventional disk drive device and controller, does not require special hardware, and is controlled by the control program executed by the central processing unit. The table shown in FIG. 5 is recorded on a disk, read into the memory 102 at system startup, and in subsequent processing, disk access is performed based on the table read in the memory.

(Flow of Reading Method) FIG. 6 is a flow chart showing the delivery processing shown in FIG. here,
A case where an image delivery request is issued from one terminal for one title will be described. Requests from other terminals are incorporated by decision flows 213 and 238 in the figure. First, when a title delivery request is issued from the terminal 1 (200), the title information table in which the relevant information of the title is recorded and the relevant information are stored in the memory 108 in the server control unit 110 from the storage device 103 of FIG. Load (201). Since a disk device is assumed as the storage device 103, the storage device and the disk will be used synonymously in the following description. In the title information table, the number of image group GOPs that make up the title,
The storage start address of each GOP, the storage start address of the I frame information in each GOP, and the like are stored, and are loaded into the memory together with the one-terminal allocation time T, the double speed number requested by the terminal J, and the like.

Next, the double speed number S (1) is read (20
2). Here, if S (1) is 1, normal speed reproduction, S
If (1) is 2 or more, it means special reproduction. Where B
define cnt as the counter of the image group GOP,
Initialize (203). In the subsequent process, S (1) is determined (204). If S (1) is 1, it is regarded as a request for normal speed reproduction, and the first GOP, that is, GOP (Bcn
The head accesses the target address of the disk drive storing t)) (205). After the reading of the data for GOP (206) is completed, the counter Bcnt is incremented by 1 (207). Next, the second GOP is read in the moving process 208 and the reading process 209, and Bcnt is incremented by 1 again in the process 210.

Here, the value of Bcnt is compared with the GOP number N of the title to be read (211). Bcnt is
If it exceeds N, the number of GOPs constituting the title has been exceeded, so this processing is ended and the next processing is executed. If Bcnt is less than N, it is determined whether a delivery stop command has been issued (212). If a delivery stop command is issued here, this processing is stopped and the next processing is executed. If the delivery is not canceled, it is investigated whether or not the title is accessed from another terminal (213), and if the access from another terminal is generated, the arbitration process 214 is executed. The arbitration process determines the priority of the delivery request issued from each terminal, and particularly when there is a risk of impairing the continuous display of images,
In order to guarantee continuous display, it is conceivable to perform processing such as reading the target data in advance. After the arbitration process or when there is no access from another terminal, S (1) is read as the next process (215). After this,
The process 204 is repeated again. Processing 2 in FIG.
The part from 05 to processing 211 is the execution part 250 of the control software which is performing the GOP reading processing during normal reproduction.
It is.

Next, a case where special reproduction is requested will be described. In the figure, if S (1) is greater than 1 (however, a natural number), the process proceeds to the process 230 after the process 204. The storage position AI (GOP (B
(cnt)) is accessed (230). Next, the I frame data is read (231) and Bcnt is set to the previous B
S (1), that is, the double speed number requested by the terminal is added to cnt (232). Next, the moving process 233 to the storage location of the second I frame, the reading of the I frame 23
4 and increment Bcnt by double speed (23
5) After that, Bcnt and N are compared, and if the number of GOPs forming the title is exceeded, the process proceeds to the next process.
If the delivery stop command 237 is issued, proceed to the next process. If not, it is judged whether or not there is access to the title from another terminal (238), and if there is, an arbitration process 23.
9 is carried out, and if there is no, the value of S (1) is confirmed (24
0). After that, the process is continued from the process 204 again.

By performing the above-described processing, special reproduction can be realized with smoother quality display than before without reducing the number of delivery channels during special reproduction.

<Embodiment 2> FIG. 7 shows that two I-frames are read within one GOP read time by increasing the number of frames constituting one GOP and decreasing the capacity of the I-frame from the reference state. As a result, smooth special reproduction is realized without reducing the number of delivery channels and compared with the conventional one.

(Device Configuration) The device configuration for realizing the second embodiment is the same as that of FIG. 1, and the control software in the delivery server 100 and the display method in the terminal 150 are changed as described below. Can respond.

(Data Structure) The data structure used in the second embodiment can also be dealt with by using the MPEG format already described in FIG.

(Storage Method, Related Information Table) The method of storing data on the disk can also be handled by adopting the method shown in FIG. The structure of the table indicating the data storage information is the same as in FIG. In FIG. 7, as in FIG. 3, terminal A and terminal C perform normal speed reproduction,
The terminal B shows the case where double speed reproduction is performed.

(Reading Method) The reading time from the disk of FIG. 7 is shown in FIG. FIG. 8 (a) is the same as FIG. 4 (a), and shows the time relationship in the case of reading without any special measures. FIG. 8B is a diagram showing the relationship of the read processing time according to the second embodiment. GO
It is assumed that the number of constituent elements of P is increased to 22 to 23 frames, which is 1.5 times as large as that in the reference state, and the maximum capacity of the I frame is half the reference state. In FIG. 8 (b), the read time required for normal speed reproduction is 20 ms for the overhead time and 30 ms for the GOP, for a total of 50 ms. On the other hand, at the time of special reproduction, two overhead times and the read processing of the I frame which requires 5 milliseconds per one sheet require a total of 50 milliseconds. Data is stored on the disk by the method shown in FIG. 2 as in the first embodiment. Therefore, the overhead time that occurs when the second I frame is read during special reproduction is generally shorter than 20 milliseconds. In any case, the read processing time required for special reproduction can be set to be equal to or shorter than the processing time for reading from the disk by one process for one terminal during normal reproduction.

Similar to the first embodiment, with respect to the reference state shown in FIG. 8A, how often I frames can be read out without reducing the number of delivery channels during special reproduction. explain. Rewriting Equation 1, the reduction rate R of the number of delivery channels is given by R = (OH time + I frame read time) × (GOP frame number / 15 frames) × α / (OH time + GOP read time) (Equation 1) To be Substituting the numerical values in the case of the second embodiment shown in FIG. 7, R = (20 ms + 5 ms) × (22.5 frames / 15 frames) × α / (20 ms + 30 ms) = (3/4) α (formula 3) is obtained. Here, from the condition of R ≦ 1, an evaluation factor indicating the smoothness of display, that is, α, which is the I-frame display frequency compared with the case of reproducing one I-frame for 15 frames in the related art, is α ≦ 4/3. That is, one I frame is reproduced and displayed for every 12 frames, which is 3/4 times the maximum of 15 frames as compared with the reference state, and the smoothness of image display during special reproduction can be improved.

In this embodiment, since the number of GOP frames is larger than the reference number of frames, the display time interval of reference image frames during fast-forward reproduction is set smaller than the display time interval of GOP during normal reproduction. By itself, the time interval of the reproduced I-frame does not become shorter than that during normal reproduction of the frame composed of the image frames of the reference number of frames. In order to perform double speed reproduction at a time interval shorter than the time interval for normal reproduction of frames of the reference number of frames, the reference frame number and the actual GOP are set to the normal reproduction time interval of the actual GOP. It is necessary to display at a time interval shorter than a time interval smaller than the ratio (frame number ratio) to the number of frames. In this embodiment, FIG.
As can be seen from the above, the average time required to read each I frame is 25 ms, and the average time required to read 1 GOP is 50 ms. The number of frames of 1 GOP is 22 to 23, and the standard number of frames is 1.
It is 5 frames. In order to be able to reproduce the I frame in a shorter time than the reproduction time when the GOP having the reference number of frames is normally reproduced, 1 GOP read time (50 ms in this example) is set to the actual number of GOP frames. Each I-frame has to be read out on average in a time shorter than a limit value (33.3 ms) divided by a ratio (22.5 / 15) to the number of reference frames. In the present example, the average value of the read time of the I frame is 25 ms, which satisfies this condition. This limit value 33.
Ratio of 3ms to I-frame read time 25ms 4 /
No. 3 gives a limit not to increase the exclusive time of the disk storage device when playing back only the I frame at the fast forward. It should be noted that in this embodiment, the reproduction interval of the I frame is only 3/4 times as long as the normal reproduction and does not become 1/2.
Strictly speaking, it is not the double speed reproduction, but for the sake of simplicity, the reproduction in this embodiment may be called the double speed reproduction. Of course, if the disk access overhead is reduced, the double speed in the accurate sense can be realized.

(Expanded Display Method) Display on the terminal will be described again with reference to FIG. In FIG. 7, a double speed is realized with respect to the terminal B by extending and displaying two I frames within the time for extending and displaying 1 GOP on the terminal side, that is, within the 1 GOP average playback display time. Since the time for displaying one frame at the normal speed at the terminal is 1/30 second, it is possible to switch the display of the I frame during the special reproduction by measuring the time on the terminal side.
In the case of the present embodiment, two I frames are reproduced for 1 GOP, which has the feature that smooth special reproduction is possible. In FIG. 7, the decompression process of the read I frame information and the display process on the terminal are performed in parallel.
This is because the buffer 153 is used on two sides, and the display is made on the other side while being expanded, so that the reproduction display can be performed without interrupting the display.

(Flow of Reading Method) The control on the server side for executing the second embodiment shown in FIG. 7 can be dealt with by partially modifying the processing shown in FIG. In FIG. 6, in the process 250 for normal speed reproduction,
This can be realized by completing one GOP reading process from process 205 to process 207 and shifting to the next process 211. The processing 251 in the case of special reproduction is equivalent to the processing itself for reading out two I frames, and the increment of Bcnt is made to correspond so that the I frame number to be read becomes the number shown in FIG. 9 described later. It can be realized by going. For example, in the case of double speed reproduction, it is possible to deal with the place where S (1) is incremented by incrementing S (1) / 2. I frame numbers to be read at other double speeds will be described with reference to FIG.

(Read I-frame Number for Each Double Speed) (For Even Double Speed) FIG. 9A shows the number of the I frame to be read at the even double speed. The top row in the table is G
This is the OP read time number and represents the order of the 1 GOP average read time in FIG. 7. In the second embodiment, one GOP is read at the time of normal reproduction, but the I frame number to be read at a read time equivalent to this is described in the table.

In the case of double speed reproduction, 1, 2, 3 ,. . .
The I-frames consecutively read are read out, and two I-frames are reproduced at the time of reproducing 1 GOP at the time of normal reproduction, thereby performing display that is twice as advanced in time. Hereinafter, at the time of quadruple speed, 2, 4, 6 ,. . . In the order of, at 6x speed,
3, 6, 9 ,. . . This can be handled by reading in the order of.

The number of the I frame to be read at each even speed is given by the following equation. Now, assuming that the double speed number is a double speed, an / 2 (however, n = 1, 2, 3, 4,) with respect to the order of the reference image frame to be read (frame number n).
5 ,. . . And below the image group number that constitutes the title). In the above equation, the first term is a / 2 and the tolerance is a
It can be obtained as a general term of the arithmetic sequence of / 2. According to the formula, (first term) + (term number -1) x (tolerance),
Substituting various values, (a / 2) + (n-1) × (a /
2) = an / 2.

(In the case of odd-numbered speed) Next, the display operation when the odd-numbered speed fast-forward reproduction is requested will be described. FIG.
(B) is an example of I to be read in the case of 3 × speed reproduction.
6 is a table showing frame numbers. Two methods are shown here. This is classified by the difference between the reference image frame number read first and the number read next.
Note that the numerical value added below each good start number is the difference between the respective frame numbers, and reading is performed so that the difference is 3 and the difference is 4 so that averaging is achieved. In the table, since the appearance of the image is the same in the method 1 and the method 2 once the triple speed reproduction is started, either one is set in the actual operation.

The I-frame number to be read in the case of odd multiple speed is given by the following formula. Assuming that the double speed number is b double speed, the order of frame numbers is n = 1, 3, 5,
7,. . . If the odd number is less than or equal to the maximum image group number that composes the title, {(b-1) / 2 + (n
−1) × b / 2}, n = 2, 4, 6, 8 ,. . . If it is an even number and is less than or equal to the maximum image group number that configures the title, it is given by bn / 2.

The former equation is an arithmetic progression with the first term being (b-1) / 2 and the tolerance being b. Where n = 1, 3, 5,
7,. . . With p = 1, 2, 3, 4 ,. . . To restate it as a general term of n = (b-1) / 2 + (p-1)
Xb. The relationship between p and the frame number n introduced here is also an arithmetic sequence of n = 1 + 2 × (p−1), and this is solved for p to obtain (b−1) / 2 + (p−1) × b
It can be obtained by substituting into.

The latter equation is an arithmetic progression with the first term being b and the tolerance being b. Here, n = 2, 4, 6, 8 ,. . .
Q = 1, 2, 3, 4 ,. . . When expressed as a general term of n = b + (q-1) × b, and solving for b,
q = n / 2 is obtained. Using this q, it can be obtained.

(Read I-frame number in the first embodiment)
Similarly, in the case of the first embodiment described above, FIG.
, I included in each GOP read time number
This can be handled by skipping the first I-frame of the frames. That is, in the first embodiment, 2 is set during normal reproduction.
Since one GOP is read and two I frames are read during special playback, for example, if the speed is 2 ×, the I frame numbers are 1, 3, 5 ,. . . , Or 2, 4, 6 ,. . . It should be read in the order of. The same applies to odd speeds.

<Third Embodiment> A method of using a plurality of disks in parallel is conceivable in order to store a large amount of information such as moving image information.

The device configuration is similar to that of FIG.
It is possible for the specific configuration of No. 3 to have a plurality of disks and for the storage control device 123 to control a plurality of disks. The control of a plurality of disks can be handled by applying the control method used in the disk array device. FIG. 11 shows a method of storing information in each disk device when four disk devices are simultaneously operated. As shown in FIG. 11A, three titles are divided by GOP. Figure 11 shows the storage on the disk.
As shown in (b), the same title is cyclically stored in the adjacent disks in order. After the storage of one title is completed, the storage of the second title is started. FIG.
In 1 (b), the head GOP of each title, that is,
D1-1, D2-1, and D3-1 are all stored on the leftmost disc, but in actual operation, storage of the title is started from the disc next to the disc on which the previous title has been stored. It doesn't matter. The difference from the storage method shown in FIG. 2 is that consecutive GOPs of the title.
Means that they do not exist in the same disc. Therefore, in the method of continuously reading out two consecutive GOPs as shown in the third embodiment, D1-1 and D1-2
Can be dealt with by storing them in continuous areas of the same disk.

FIG. 11C corresponds to FIG. 2D, and in the case of four disks, D1-1 and D1.
-5 is stored in the continuous area.

FIG. 12 is a diagram showing a reading method from a disc and a reproducing method at a terminal when the storage shown in FIG. 11 is performed. In FIG. 12, the case where the terminal B is performing 5 × speed reproduction is described. 1 GOP as shown
Within the average read time, if the information just distributed and stored in each disk is read, it is possible to read four I frames. In FIG. 12, D2-1 to D2
-4 when 4 I-frames are being read out, 1G
If a plurality of I frames can be read within the OP average read time, the same effect can be obtained. Depending on the double speed number, the effect of improving the reading efficiency due to the simultaneous operation of a plurality of disks may not appear. This occurs especially at low speeds. Taking the case of FIG. 12 as an example, for example, 2
For double speed. Since only two I-frames D1-1 and D1-2 have meaning, in this case, D1-1 and D1-1 from the disk 1 and D1- from the disk 2 are used.
As shown in 2, D1-2, the same I frame is read twice or once each, and each is displayed on the terminal for a playback time corresponding to 1/2 GOP. Also in the cases of FIGS. 11 and 12, the increase / decrease in the number of delivery channels during the normal reproduction and the special reproduction, and the smoothness of the image during the special reproduction are, on average, the same as those of the second and third embodiments. . Of course, the total number of delivery channels itself is 1 because the number of disks operating simultaneously has increased.
It increases compared to the case of a stand.

With the above storage method, the same effect as in the first embodiment can be obtained in this embodiment as well.

[0088]

According to the present invention, in a system for delivering continuous information such as moving image information from a storage device provided in a server to a large number of terminals, special reproduction such as fast-forward reproduction and rewind reproduction is required. In such a case, it is possible to reproduce smooth and well-connected images without increasing the time occupied by this storage device compared to the time of normal reproduction.

[Brief description of drawings]

FIG. 1 is a diagram showing an example of a device configuration according to an embodiment of the present invention.

FIG. 2 is a diagram illustrating a method of continuously storing image information on a magnetic disk and simultaneous delivery to a plurality of terminals.

FIG. 3 is a diagram for explaining a disk reading at the time of moving image delivery and a display method on a terminal according to the first embodiment of the present invention.

FIG. 4 is a diagram illustrating a read time of image information from the disc of the first embodiment.

FIG. 5 is a diagram showing a structure of management information regarding storage of moving image information.

FIG. 6 is a diagram illustrating a processing flow of a delivery method when a special reproduction request is made from a terminal.

FIG. 7 is a diagram for explaining a disk reading at the time of moving image delivery and a display method on a terminal according to the second embodiment of the present invention.

FIG. 8 is a diagram illustrating a read time of image information from the disc of the second embodiment.

FIG. 9 is a diagram illustrating a read frame and a display method when performing even-numbered speed reproduction and odd-numbered speed reproduction.

FIG. 10 is a diagram illustrating a data format of an image compression method used in the present invention.

FIG. 11 is a diagram illustrating a method of storing image information on a plurality of discs according to the third embodiment of this invention.

FIG. 12 is a diagram for explaining a disk reading operation at the time of moving image delivery and a display method on a terminal according to the third embodiment of the present invention.

[Explanation of symbols]

100 is a delivery server, 101 is a central processing unit, 103 is a storage device, 140 is a communication line, 150 is a terminal, 152 is a decompression device, 153 is a buffer, 154 is a display device, 16
Reference numeral 1 is a central controller.

 ─────────────────────────────────────────────────── ─── Continuation of the front page (72) Inventor Takuo Kusaba 1-280, Higashi Koikekubo, Kokubunji City, Tokyo Inside Hitachi Central Research Laboratory (72) Inventor Akihiko Oba 1-280, Higashi Koikekubo, Kokubunji City, Tokyo Hitachi, Ltd. Central Research Laboratory (72) Inventor Mitsuo Asai 1-280, Higashi Koikekubo, Kokubunji City, Tokyo Hitachi, Ltd. Central Research Center (72) Inventor Mitsuko Sato 1-280, Higashi Koikeku, Kokubunji, Tokyo Hitachi Central Research Center Co., Ltd. (72) Inventor Yoshihiro Takiyasu 1-280, Higashi Koigokubo, Kokubunji, Tokyo Inside the Central Research Laboratory, Hitachi, Ltd.

Claims (15)

[Claims] 1. A delivery server for storing a plurality of moving image information, and delivering the moving image information requested by the client to be delivered from the stored plurality of moving image information to the client via a communication line. However, the delivery server includes a storage device that holds the plurality of moving image information and a control device that controls reading of the moving image information from the storage device, and each moving image information includes a plurality of image groups, Each image group is
Corresponding to each one of the image frames, corresponding to a plurality of reference image frames consisting of information obtained by spatially compressing the information of the image frame, and to each subsequent image frame of the one image frame A plurality of interpolated image frames, and each interpolated image frame is temporally compressed based on the reference image frame with respect to the plurality of image frames from the reference image frame to the corresponding one image frame. In addition, in a moving image delivery system consisting of an interpolation image frame obtained by spatially compressing the information of the corresponding image frame, fast forward reproduction or rewinding of any moving image information from the client. If playback is requested, skip the interpolated image frames of the image groups with different moving image information. A plurality of reference image frames of these image groups are sequentially read from the storage device, and the plurality of read reference image frames are sequentially delivered to the client. The plurality of delivered reference image frames are reproduced at a reproduction time interval smaller than the normal reproduction time of, and the reproduction time interval has a ratio of the time interval and the normal reproduction time of each image group to each moving image information. A moving image that is determined to be equal to or greater than the ratio of the average time required to read a plurality of reference image frames belonging to the plurality of image groups to the average time required to read each image group from the storage device. Image reproduction method. 2. In the storage, a plurality of image frame information in the same image group is divided into pairs of two image frames, and two pieces of image frame information belonging to the same pair are consecutively stored in the storage device. When the client requests the fast-forward reproduction of certain moving image information, it is stored in an accessible storage position, and when the client requests the fast-forward reproduction of the certain moving image information, the two reference images included in the moving image information belong to the same pair. 2. The moving image reproducing method according to claim 1, wherein two reference image frames belonging to one image group are read out successively. 3. The storage device has one disk device having a plurality of disk surfaces, and when storing, a plurality of image frames belonging to the same image group are stored on adjacent disk surfaces, and the same cylinder is full. After that, if there are subsequent image frames in the image group, those subsequent image frames are stored in the adjacent cylinders, and if there are no subsequent image frames, the image frames in the subsequent image group are stored in the adjacent cylinders. The moving image reproducing method according to claim 1, wherein the moving image is stored in a cylinder. 4. The moving image according to claim 1, wherein the storage device has a plurality of disk devices, and the plurality of moving image information is cyclically and sequentially stored in different disk devices in units of image groups during the storage. Image reproduction method. 5. A reading time required when reading out a plurality of reference image frames belonging to a plurality of image groups of each moving image information in succession, and a time required to read out each image group from the storage device, The prefix overhead time from the issuance of the read command for the image information to be read to the storage device to the actual start of reading, the read time from the start of reading to the end of reading, and after the end of the read processing 2. The moving image reproducing method according to claim 1, wherein the moving image reproducing method comprises a total of post-position overhead times until a subsequent read command can be accepted. 6. An MPEG (Moving) as an image compression / decompression method.
Picture coding Expert Group), which is a user-defined area in the sequence header information that is located at the beginning of the image group defined by the standard, and stores management information such as the storage location, image group capacity, and reference image frame capacity related to the image group. The moving image information reproducing method according to claim 1, wherein 7. A delivery server which stores a plurality of moving image information, and delivers the moving image information requested by the client to be delivered from the stored plurality of moving image information to the client via a communication line. However, the delivery server includes a storage device that holds the plurality of moving image information and a control device that controls reading of the moving image information from the storage device, and each moving image information includes a plurality of image groups, Each image group is
Corresponding to each one of the image frames, corresponding to a plurality of reference image frames consisting of information obtained by spatially compressing the information of the image frame, and to each subsequent image frame of the one image frame A plurality of interpolated image frames, and each interpolated image frame is temporally compressed based on the reference image frame with respect to the plurality of image frames from the reference image frame to the corresponding one image frame. In addition, in a moving image delivery system including an interpolated image frame obtained by spatially compressing the information of the corresponding image frame, each of a plurality of image groups of each moving image information is configured. As an image frame, an image group composed of a predetermined number of image frames larger than a predetermined reference number of image frames When a fast forward playback or rewind playback of any moving image information is requested by the client, multiple criteria of these moving image information are skipped while skipping interpolation image frames of different moving image information. The image frames are sequentially read from the storage device, the plurality of read reference image frames are sequentially delivered to the client, and the client normally reproduces the image group including the reference frame number of image frames. The plurality of delivered reference image frames are played back at a playback time interval smaller than the normal playback time at the time, and the playback time interval is a ratio of the time interval and the normal playback time of the image group of the reference number of frames. , Each of a plurality of reference image frames belonging to a plurality of image groups of each moving image information A moving image reproducing method having a value that is equal to or more than a ratio of an average of the time required for reading and an average of the time required for reading each image group of the moving image information from the storage device. 8. The average ratio of the capacities of reference image frames in the image group to the capacities of the image groups is 4
The moving image reproducing method according to claim 7, which is less than or equal to one-tenth. 9. A read time required when two reference image frames belonging to a plurality of image groups of each moving image information are consecutively read, and a time required to read each image group from the storage device are: The prefix overhead time from the issuance of the read command for the image information to be read to the storage device to the actual start of reading, the read time from the start of reading to the end of reading, and after the end of the read processing 8. The moving image reproducing method according to claim 7, comprising the sum of post-position overhead times until a subsequent read command can be accepted. 10. A reference image frame to be read when reproduction is performed at an even multiple speed equal to or higher than double speed, in which double speed numbers are a, n are 1, 2, 3, 4 ,. . . Is a natural number and is less than or equal to the maximum image group number of the target moving image, (an / 2)
The moving image information reproducing method according to claim 1, wherein the reference frames in the th image group are sequentially read out. 11. When a double speed number is set as b as a reference frame to be read out when reproduction is performed at an odd double speed of 2 × or more, n is 1, 3, 5, 7 ,. . . Is an odd number and less than or equal to the maximum image group number of the target moving image, {(b-1)
/ 2 + (n-1) b / 2} reference frames in the image group are sequentially read, and n is 2, 4, 6, 8 ,. . . 2. The moving picture information reproducing method according to claim 1, wherein the reference frame in the (bn / 2) th picture group is sequentially read out when it is an even number and less than or equal to the maximum picture group number of the target moving picture. 12. When the number of double speeds is b, and n is 1, 3, 5, 7 ,. . . Is an odd number and less than or equal to the maximum image group number of the target moving image, {(b + 1)
/ 2 + (n-1) b / 2} reference frames in the image group are sequentially read, and n is 2, 4, 6, 8 ,. . . 2. The moving image information reproducing method according to claim 1, wherein the reference frames in the (bn / 2) th image group are sequentially read out when the number is even and the maximum image group number of the target moving image or less. 13. A delivery server, which stores a plurality of moving image information, and delivers the moving image information requested by the client to be delivered from the stored plurality of moving image information to the client via a communication line. However, the delivery server includes a storage device that holds the plurality of moving image information and a control device that controls reading of the moving image information from the storage device, and each moving image information includes a plurality of image groups, Each image group corresponds to one of the image frames, and a plurality of reference image frames each of which includes information obtained by spatially compressing the information of the image frame and a subsequent one of the image frames. A plurality of interpolated image frames corresponding to one image frame, and each interpolated image frame includes a plurality of interpolated image frames from the reference image frame to the corresponding one image frame. A moving image composed of an interpolated image frame obtained by temporally compressing an image frame based on the reference image frame, and further spatially compressing information of the corresponding image frame. In the delivery system, the storage device has one disk device having a plurality of disk surfaces, and stores a plurality of image groups belonging to each moving image information so as to be located in the same track of the disk storage device. , When a fast forward reproduction or a rewind reproduction of any moving image information is requested by the client, a plurality of reference image frames belonging to different image groups of the moving image information are sequentially read from the storage device and read out. The plurality of reference image frames are delivered to the client, and the plurality of delivered reference image frames are delivered to the client. , A moving image reproduction method of reproducing at a time interval smaller than the normal reproduction time of each image group. 14. At the time of the storage, a plurality of image frames belonging to the same image group are stored on adjacent disk surfaces, and after the same cylinder is full, if there are subsequent image frames of the image group, those image frames are stored. 14. The moving image reproducing method according to claim 13, wherein the subsequent image frames are stored in adjacent cylinders, and when there are no subsequent image frames, the image frames of the subsequent image group are stored in the adjacent cylinders. 15. The moving image according to claim 13, wherein said storage device has a plurality of disk devices, and at the time of said storage, said plurality of moving image information are cyclically stored in different disk devices in units of image groups. Image reproduction method.