JP3177366B2 - High-speed image playback system - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION The present invention relates to an MPEG (Moti)
on Picture Experts Group)
Moving image compression technology (IS) standardized by ISO
The present invention relates to a high-speed image reproduction system for reproducing a medium on which image data compressed by O. 11172) is recorded at a high speed.

[0002]

2. Description of the Related Art MPEG (Motion Picture)
The image data compressed by the moving image compression technology (ISO 11172) standardized by ISO called e Experts Group) includes an I Picture compressed and encoded in an image unit and an image located in the past in time. P P compressed using the correlation of
Icture and I located in the past and future in time
Picture or BPic, which is an image compressed by taking correlation and difference from P Picture
cure. (These video compression techniques are currently being standardized by ISO.
Among them, an algorithm for compressing audio and video at a transmission rate of 1.5 MBPS is called MPEG1, and is an ISO.
It is standardized as 11172. The following procedure is used in the moving image compression by the MPEG system. This is shown in FIG.

In the moving image compression by the MPEG system, first, an I Picture (21) which is an image to be compressed in an image unit is compressed from a plurality of images. This is a compression method capable of restoring an image using only the compressed data. As a compression method, in the case of MPEG, this image data is divided into a certain area (8 × 8 pixels), and an orthogonal transformation such as a two-dimensional discrete cosine transformation is performed on the area to obtain a frequency domain. The compression is performed by making use of the fact that the human visual characteristics are insensitive to high frequencies, performing weighting, assigning many codes to low frequency regions, and reducing code allocation in high frequency regions. . This is performed by means for providing encoding accuracy called quantization. As a result, it is said that an image having no practical problem can be obtained as an I Picture if the data amount is about a fraction. And
The other temporally succeeding images 22 and 23 are compressed by referring to the I Picture image data thus compressed.

Next, a compression method by reference will be described. Normally, images compressed by this method are classified into images that are referred to only from past images and images that are referred to from past and future images. As an encoding procedure, first, an I Picture is encoded, and then a P Picture, which is an image to be compressed and decoded with reference to a past image, is encoded. Then, B Picture, which is an image to be compressed and decoded by referring to past and future images, is encoded. This image is usually between the two images in time, and the I Picture and P Picture
It gives less code than re. FIG. 3 shows an example of such an image sequence. Here, image I is I Pic
cure, image P is P Picture, image B is B
Picture. As described above, there are three types of images in the image compressed by this method.
The image is one out of several to several tens of images, the P Picture which is referenced only from the past image data is one out of two or three, and the others are images which are referenced from the past and the future. It will be B Picture. However, the number of these images is not determined, but is determined by the state of the encoder, the amount of information generated so far by encoding, and the like.

[0005] P Picture, B Picture
Is performed by taking a correlation with another image. Here, each image is divided into predetermined regions, a region having a pixel value closest to the divided region data is searched for from an image having a correlation, and a difference between the positions is encoded. This is called a motion vector. However, the region selected in this way is less likely to have the same value as the region to be compressed. Therefore, it is necessary to encode the difference data of the pixel value between the reference area and the reference area.
At the time of encoding this data, compression is performed by the same orthogonal transformation means and quantization means as used in I Picture.

The decoding and reproduction of moving image data compressed by this compression method will be described. FIG. 7 shows a conventional MPE.
1 shows an example of a G reproduction system. During playback,
First, the I Picture is decoded to form an image. The MPEG data recorded on the recording medium 701 is
It is read by the recording medium reading means 702. As this recording medium, optical, magnetic disk, optical, magnetic tape, I
Various recording media such as a C card can be used. The MPEG data read from the recording medium 701 by the recording medium reading means 702 is decoded by the variable length decoding means 713. These data are
The motion vector data by the data switching means 703;
It will be divided into orthogonal transform information.
Since all the ure data are orthogonal transform information, all of the data in this case is sent to the inverse quantization means 704, and inverse quantization, which is the reverse procedure of quantization, is performed. Then, it is sent to the orthogonal transformation information decoding means 705, and each pixel data of the image is obtained. In the case of I Picture, since there is no data generated from the motion vector as described above, decoding of the image is completed by passing through the adder as it is. This decrypted IP picture is stored in the image memory 70.
7 is stored.

[0007] Based on this image, another compressed image (P Picture, B Picture) is restored. In this decoding, first, an image to be restored is treated as a plurality of regions similar to those at the time of encoding, and decoding is performed on each of them. The data for each area is input to the data switching means 703 and separated into motion vector data and difference data.

[0008] The data of the motion vector is 70
8 or 710 in the figure to perform motion compensation. The difference data is input to an inverse quantization unit 704 that performs inverse transform of quantization, and the result is input to an orthogonal information decoding unit 705, and the difference data is decoded. The area is decoded by adding these two output data by the adding means 706. Then, by decoding each area in the image, decoding of one image is completed. Among the decoded images, P Pictures, which are images referred to by other images, are sequentially stored in the image memory 709 and used for restoring other images.

In the MPEG system, I Pic
Since decoding is performed in the order of “ture”, “P Picture”, and “B Picture”, the order of decoding is not necessarily the order of reproduction. For this reason, the decoded image data is stored
7, 709, etc., and may be output as an image after referring to decoding of other images. Therefore, means for switching these image data is required. This switching is performed by the image switching means 712.

[0010] As described above, in a reproduction apparatus of digital moving image data compressed by correlating an image compressed in an image unit with the image data and encoding the difference data, first, several images are reproduced. Compressed image data (I Pictur) within an image unit
e) is first decoded. Based on this image data, using the correlation data obtained at the time of encoding, an image closest to the area of the reproduced image is extracted from this image with respect to a certain area data of the reproduced image. Then, a difference between the image area data and the actual image data is calculated at the time of encoding, and the data is restored and added to obtain a reproduced image (P Pic).
tuer, B Picture) is obtained.

As described above, in the data reproducing apparatus, most of the images require the I Picture that appears once between several to several tens of images,
When reproducing an image other than the image, the IP
It is necessary to first form an image and to form an image with reference to it. Calculation of these correlations includes motion compensation,
A means called orthogonal transformation of difference information is used. As the orthogonal transform means, a two-dimensional discrete cosine transform is performed in the MPEG system, and a huge amount of calculation is required to obtain the orthogonal cosine transform. For this reason, when performing fast-forward high-speed playback, the amount of calculation for obtaining all image data becomes enormous, and as a result, calculation cannot be performed at a normal decoder speed. For this reason, I
Fast forward playback has been realized by playing back only the Picture. This method is typically described in JP-A-63-310293, in which compressed image data compressed within an image unit, that is, IP in the present invention.
A technique for decoding and reproducing only the title is devised. However, this reproduction method has a drawback that only about three images can be obtained at most per second, and only a very unnatural and discontinuous fast-moving moving image can be obtained as a moving image. .

[0012]

In the conventional high-speed reproduction method described above, since only the I Picture is reproduced, the motion in the image becomes unnatural or discontinuous. In the case of data having a small number of pictures, there is a further problem that the image can be obtained only in an unnatural or discontinuous and awkward state.

The present invention has been made in view of such a problem, and a high-speed reproduction system capable of obtaining a smoother moving image by simultaneously decoding not only a self-compressed image but also other images. Is provided.

[0014]

SUMMARY OF THE INVENTION In order to solve such a problem, the present invention provides a recording medium which records MPEG data composed of I picture data, P picture data and B picture data in a predetermined order. A reproducing means for reproducing at a normal reproducing speed or a speed of N times (N> = 2); and decoding a difference component of the encoded data of each picture,
Orthogonal transformation information decoding means for outputting independent image data of an I picture and differential image data of a P picture and a B picture, and an image memory for sequentially updating and storing image data formed corresponding to the I picture and the P picture If the detected P-picture data and moving-out vector data of the B in picture data, and a data reading means for reading the image data from the image memory in accordance with the motion vector data corresponding to the image data to be formed, is read Adding means for adding the obtained image data and the difference image data to output image data; I picture image data obtained from the orthogonal transformation information decoding means; and P picture image data obtained from the image memory. The image data of the B picture obtained by the adding means is
In an image reproducing system provided with a selecting means for selecting the image data in accordance with the output order of the image data, during the reproduction period of the P picture data and the B picture data in conjunction with the N-times speed reproduction operation of the reproducing means. A control means for selectively performing only decoding of a low-order area with respect to the orthogonal information decoding means.

[0015]

According to the present invention, in a reproduction system for reproducing data compressed by the MPEG method, when performing high-speed reproduction, only low-frequency components of orthogonal information are decoded between I-pictures and composed of motion-compensated data. The interpolated image is interpolated.

[0016]

BRIEF DESCRIPTION OF THE DRAWINGS FIG.
FIG. 1 shows a configuration diagram of the present invention. A reading means 10 for enabling a disc player or a tape player capable of fast-forward reproduction from a recording medium 101 on which MPEG data is recorded.
2, the image data compressed and encoded is read. As the recording medium on which the MPEG data is recorded, any medium capable of high-speed reading may be used, and various recording media capable of high-speed reading, such as an optical disk, a magnetic disk, an optical disk, and a magnetic tape, may be used. . The data read from these recording media is subjected to variable-length decoding by the variable-length decoding means 113, and it is necessary to determine whether to switch the decoding of the data as a motion vector or as orthogonal transformation information based on the decoding result. There is. This output data is sent to the data switching means 103. At the same time, detection of a header that determines the type of image in the data is performed at 108
Is performed by the header detecting means.

In the case of normal reproduction, first, in this apparatus, I
Picture is restored. All the compressed data of this image undergoes inverse quantization via an inverse quantizer 104 which performs inverse transform of quantization, and is input to an orthogonal transform information decoder 105 to be decoded. The IPi restored in this way
The result data is output as an image, and is stored in the image memory 106 as data for restoring the next image.
Is input to

Next, the P Picture is restored using the I Picture data. The input data is sorted by the data switching unit 103. The input data includes data relating to motion information between the regions and data relating to differences in that case. Data relating to the motion information is sent to the motion compensating means 107. Here, an optimal area is extracted from the previous image data using the motion information. Data regarding the difference is input to the inverse quantizer 104, and the output is sent to the orthogonal transform information decoder 105. Here, M
In the case of PEG1, an inverse DCT operation is performed. Also,
Here, other transforms, for example, a Haar transform, a Hadamard transform, a Fourier transform, a WAVELET transform or the like frequency transform from space may be used. As a result, the value of the difference between the regions is restored, and the image data is restored by adding these two data by the adding means 108. This operation is performed for the entire area of the target image, and the entire image is restored. The image restored by referring to the I Picture is stored in the image memory 110.

Then, the P Pictu that follows in time is
re will be sequentially decoded using this stored P Picture. Next, at the time of decoding the B Picture, the image is decoded using the two image data stored in the image memories 106 and 110 in the same manner as in the case of the P Picture. However, in this case, since data of two images is extracted, the distribution ratio of the pixel values is determined by the motion compensation 107, 111 or the adder 108.

In this MPEG system, I Pic
Since decoding is performed in the order of “ture”, “P Picture”, and “B Picture”, the order of decoding is not necessarily the order of reproduction. For this reason, the decoded image data is
6, 110, etc., and may be output as an image after referring to decoding of other images. Therefore, means for switching these image data is required. This switching is performed by the image switching unit 104.

Next, the operation at the time of fast forward reproduction will be described.
First, high-speed reading from a recording medium is performed. The faster the reproduction speed, the faster the data must be read. The control means 109 issues an instruction to the orthogonal transformation information decoding means 105 in accordance with the fast-forward speed, whereby the fast-forward reproduction is performed. The contents of this instruction will be described below.

First, since I Picture is a reference for all images, it decodes all information. And P Pictur
For e and BPicture, the control unit 109 specifies the ratio of the orthogonal transform information to be decoded to the orthogonal transform information decoding unit 105 according to the reproduction speed. When the reproduction speed is relatively slow, only the low-order data, which is part of the orthogonal transform information, which is easily restored, is restored. The ratio of the lower-order data is determined by the speed of high-speed reproduction and the calculation time required for decoding the data. In many cases, complicated calculations such as two-dimensional cosine transform (DCT transform) and frequency transform are used to encode the difference data, and this requires a large amount of calculation and a long calculation time. In the case of high-speed reproduction, for example, when triple-speed high-speed reproduction is required, the orthogonal information must be converted in one-third the time of normal reproduction. A component that can be decoded by one third is determined, and decoding is performed. At that stage, for example, half the information in the case of normal playback, or quarter information,
Alternatively, it is changed according to the speed in the form of information of only the DC component and according to the performance of the decoder. The operation required to decode the most orthogonal transform information in a range where the time required for decoding is smaller than one times the double speed is selected, and the control means 109 designates the orthogonal transform information decoding means 105. Become. On the other hand, in the process of retrieving the data of the image area by the motion compensating means, it is sufficient to change the read address of the memory, and the process can be performed at a much higher speed than the decoding of the orthogonal transform information.
As a result, these images can be formed in a very short time. As a result, although the image quality is slightly lower than the officially reproduced image, a sufficient image for high-speed reproduction is obtained.
It can be provided as an image other than the Picture.

[0023]

[0024]

Next, a description will be given of a case where fast-forward playback is performed. For example, in the case of double-speed reproduction, as a motion of the entire decoder, reading from a recording medium is performed at twice the normal reproduction. Then, in the case of double speed reproduction, it is desirable to reproduce half of the images in the image sequence.
In this case, first, the I Picture is decoded. Then, the P Picture which refers to this I Picture
re is restored. However, there is a case where it is impossible to restore the image using the entire information of the image as the speed of the decoder, so that only the reference or the low-order information of the orthogonal transform information can be reproduced at high speed. The difference data is decoded using only the components, and the image is restored. Therefore, this image has a lower image quality than the image that is actually finally restored, and is a mosaic image in which the corresponding portions of the I Picture are pasted together. However, in the case of fast-forward playback, it is sufficient that the atmosphere of the image is transmitted to the user, and thus this image is considered to be sufficient. Next, the difference data is decoded by referring to only this image or by using only components that are in time for high-speed reproduction of lower-order information of the orthogonal transformation information.
Pictures are restored, and images are restored sequentially.

A description will be given of the actual situation in which the reproduction speed is varied using the above. First, in an area where the reproduction speed is close to one time, an image is reproduced while decoding a part of the difference data within the range of the capability of the decoder. However, even if all the images are reproduced, for example, if the reproduction speed is 1.2 times,
In the case of the NTSC system, only one out of thirty images out of thirty, that is, twenty-five images can be output as hardware. For this reason, the five images are output in a blurred manner. For this purpose, first, the calculation is performed by rounding the B Picture, that is, decoding is not performed on some B Pictures. When the reproduction speed reaches a region where the difference cannot be calculated by the decoder, the difference data is not decoded or only the low-order orthogonal transform information that can be decoded is decoded. And the thinning of the image is
For example, in the case of double speed, in the case of the NTSC system, thinning of 15 images is performed. This will thin out most BPictures. B Pic
The existence of “ture” itself is not referred to by other images as a reference image for decoding, and does not affect reproduction of other images. As described above, B Picture requires two image memories for restoration. For these reasons, the reproduction output of B Picture is first limited. As described above, in order to simplify the processing, a system configuration in which only the P Picture is reproduced is simple. Where P
Even when only pictures are reproduced, all P pictures need to reproduce only motion prediction. This is because all the P Pictures are reproduced by referring to the P Picture at the closest position existing in the past, and if one of the pictures is interrupted, the subsequent P Pictures cannot be reproduced. It is.

However, if it is not necessary to reproduce the subsequent P Picture, for example, if the reproduction is performed at 8 × speed, the next I Picture is reproduced from these I Pictures.
Since only one image needs to be reproduced before the Picture, the P Pi output only once is output.
There is no need to perform reproduction except for the "cure."

The pattern for reproducing and outputting these P Pictures will be described below. For example, consider the case as shown in FIG. This is because I Pict 41 in the figure
ure, 42, 43, 44, 45, and 46 are images (P Pictures) that are referenced only from the past, and the others are images (B Pictur) that are referenced from the past and the future.
e). In this case, only one image is available from 41 to 42, and three images are available from 42 to 43 and 43 to 44, and 44 to 45 and 45 are available.
To 46 have time for two images. When the image is reproduced as it is, the reproduction of the image data itself in the image reproduction stage is, for example, about 30 per second in the case of the NTSC system.
Since playback can be performed only at the display speed of a single sheet, a problem occurs in double speed playback. For this reason, by reproducing these images at a later time, image data is output at a speed suitable for reproduction. Figure 5 shows this situation.
Shown in Here, the image 53 is reproduced twice. As described above, although the time is slightly shifted, it is possible to reproduce the time sequence almost and reproduce an image other than the I Picture.

Next, consider the case of triple speed. In the case of the triple speed, similarly to the case of the double speed, readout from the recording medium at the triple speed, restoration of the I Picture, and simple restoration of the image are performed. Since these simple restorations require little time, if the I Picture can be restored at a reproduction speed of several times, it is possible to restore the images to be referenced only from all the past. is there. However, as in the case of the double speed, the number of images (P Pictures) referred to only from the past is not constant in the image sequence, so that a problem occurs at the stage of reproduction. For this reason, by reproducing these images at a later time, image data is output at a speed suitable for reproduction. This is shown in FIG. In this case, the image 65 will not be reproduced.

When considering how many times an image is to be reproduced, it is necessary first to know how many units (No) of time before the image was reproduced before the image was reproduced. Then, it is necessary to know how many units (Nd) time can be restored after the image. This is because if there is only an image unit that is equal to or less than the double speed (Sn) of the fast-forward reproduction during this time, the image reproduced next to the previously reproduced image is the image that can be restored next to the corresponding image. Will be done. Conversely, if there is an image unit that is at least the double speed (Sn) number of the fast-forward or rewind reproduction during this time, the corresponding image must be reproduced. In particular, when there are image units that are at least twice the double speed (Sn), the corresponding image is reproduced twice. Therefore, the number of times of reproduction (Dn) in this case is given by the following equation.

[0031]

(Equation 1)

As described above, when the image is reproduced before the reproduction of the image, and when it is possible to understand the position of the next reproducible image, the reproduction is performed. The number of times will be determined. As described above, in a speed region where high-speed reading from a recording medium is possible and a compressed image (I Picture) in an image unit can be restored, an image (P Picture) to be referred only from the past is used.
re) can be easily restored, so playback on that time axis,
By performing the non-reproduction determination, an image can be output at an arbitrary speed requested for fast-forwarding, and the reproduction speed can be smoothly changed from the arbitrary speed.

As described above, at the time of fast-forward playback, an image other than the image (I Picture) compressed in the image unit, for example, P Picture, B Picture
To provide a fast-forward playback image with a relatively smooth motion.

[0034]

According to the present invention, in a reproducing system for reproducing a recording medium on which image compressed data compressed by the MPEG system is recorded, the I Pictu at the time of high-speed reproduction.
P Picture, BP which are images other than re
In the reproduction of the picture, the decoding is performed by using the orthogonal transform information only partially, so that the I Picture can be reproduced.
Other images can be reproduced, and a smooth image at the time of high-speed reproduction can be provided.

[Brief description of the drawings]

FIG. 1 is a block diagram of a digital video playback device of the present invention.

FIG. 2 is an explanatory diagram showing an example of moving image compression by the MPEG method.

FIG. 3 is an explanatory diagram showing an image compression sequence.

FIG. 4 is an explanatory diagram showing an example of fast-forward playback and a form of image data compression.

FIG. 5 is an explanatory diagram showing a division of a reproduction time of image data when performing double speed reproduction.

FIG. 6 is an explanatory diagram showing a division of a reproduction time of image data when performing triple speed reproduction.

FIG. 7 is a circuit block diagram of a reproducing apparatus according to a conventional MPEG moving image compression standard.

[Explanation of symbols]

 101: recording medium 102: recording medium reading means 103: data switching means, 104: inverse quantizer 105: orthogonal transform information decoding means 106: image memory 107: motion compensation means 108: adder 109: control means, 110: image Memory 111: motion compensation means 112: image memory 113: image switching means, 114: variable length decoding means

Claims (1)

(57) [Claims] 1. A recording medium in which MPEG data composed of I picture data, P picture data and B picture data are recorded in a predetermined order, is reproduced at a normal reproduction speed or a speed of N (N> = 2) times. And decoding means for decoding the difference components of the coded data of each picture to obtain independent picture data of I picture, P picture, and B picture.
Orthogonal transform information decoding means for outputting difference image data of a picture; an image memory for sequentially updating and storing image data formed corresponding to the I picture and the P picture; the P picture data and the B picture data detects <br/> motion-out vector data in a data reading means for reading the image data from the image memory in accordance with the motion vector data corresponding to the image data to be formed, and the read image data and the difference image Adding means for adding image data to output image data; I picture image data obtained from the orthogonal transform information decoding means; P picture image data obtained from the image memory; and B picture data obtained from the adding means. Means for selecting image data of the picture in accordance with the output order of the image data. In the high-speed image reproducing system provided in each of the above, in conjunction with the N-times speed reproducing operation of the reproducing means, a low-order area is provided to the orthogonal information decoding means during the reproduction period of P picture data and B picture data. A high-speed image reproduction system characterized by including a control means for selectively performing only decoding of the image.