JPH1032830A - Re-encoding method and device for image information - Google Patents

Abstract

PROBLEM TO BE SOLVED: To improve re-encoding image quality by deciding a quantization step size of the re-encoding mode by a certain rule with respect to a quantization step size of the preceding stage encoding mode. SOLUTION: An encoding control part 13 decides a picture type at a picture type deciding part 70, based on the I/P/B information of the preceding stage encoding mode on the output of a preprocessing part. Deciding part for a quantization step size 71 selects the quantization step sizes of the re-encoding mode in every macro block, based on the Q information of the preceding stage encoding mode, the image feature value of the output of an encoding part and the encoding date respectively. In this selection process, the quantization step size Q2 for the re-encoding mode is set larger than the quantization step size Q1 for the preceding stage encoding mode, and also a Q rule is selected which is larger than the size Q1 by a multiple of natural number is satisfied. Then an image block position deciding part 72 sets the image block position of the re-encoding mode, so as to keep the image block boundary of the preceding stage encoding mode. Thus, an encoding part performs a re-encoding operation, based on the re-encoding parameter of the output of the part 13.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a re-encoding method and apparatus for encoding image information a plurality of times.
The present invention relates to a method and apparatus for re-encoding image information in digital television transmission, a digital image storage / transmission system, an image database, and the like.

[0002]

2. Description of the Related Art In recent years, applications for decoding digital image information from an encoded bit stream, re-encoding it under different encoding conditions, and transmitting or storing the information have been increasing. For example, in the field of broadcasting, the collection of image materials, the primary distribution between television stations, the secondary distribution to homes, etc., cascading transmission of digital signals while performing editing processing and encoding processing, that is, one image Is performed a plurality of times, and as the broadcasting form becomes more diversified, a transmission scheme having a higher degree of freedom, which is different from the hierarchical transmission, is becoming widespread. For video clips stored in an image database or the like and used as a library,
At the same time as the source is provided by many users, many users utilize it and transmit and store it repeatedly with editing and encoding.

[0003] When image information is encoded a plurality of times, the image information has an encoding history. In the prior art, such image information is re-encoded in a form ignoring the encoding history. Will be That is, at the time of re-encoding, processing is performed in consideration of only an independent parameter for each re-encoder, for example, only the compression ratio of the re-encoder.

The applicant of the present application has quantitatively analyzed that re-encoding ignoring such an encoding history causes significant image quality deterioration, and in view of the result, reads out the encoding parameters in the pre-encoding. Japanese Patent Application Laid-Open No. H8-111870 has already proposed that an encoding parameter adapted to this is determined, and image information is re-encoded using the determined encoding parameter.

[0005]

In the proposed technique of the applicant of the present invention, the quantization step size at the time of re-encoding is set to a value having a high affinity in consideration of the quantization step size of the pre-encoding. However, even if re-encoding is performed using the quantization step size determined by such a method, sufficient image quality improvement cannot be obtained.

SUMMARY OF THE INVENTION Accordingly, an object of the present invention is to provide a method and apparatus for re-encoding image information which can improve the image quality of an image having an encoding history.

[0007]

According to the present invention SUMMARY OF], a method of re-encoding image information with encoded history, at the time of re-encoding the quantized step size Q ₁ in the front stage coding The quantization step size Q ₂ is determined so that Q ₂ ≧ Q ₁ and Q ₂ = n × Q ₁ (where n is a natural number), and input using the determined quantization step size Q ₂ . An image information re-encoding method for re-encoding image information is provided.

Further, according to the present invention, there is provided an apparatus for re-encoding image information having an encoding history, wherein a quantization step size Q ₁ in re-encoding is compared with a quantization step size Q ₁ in pre-encoding. ₂ with Q ₂ ≧ Q ₁ and Q ₂ = n
× Q ₁ (where n is a natural number), and re-encoding means for re-encoding input image information using the quantization step size Q ₂ determined by the determination means. And a device for re-encoding image information comprising:

The quantization step size Q ₂ at the time of re-encoding is compared with the quantization step size Q ₁ at the time of the _first- stage encoding.
However, by using a rule that satisfies Q ₂ ≧ Q ₁ and Q ₂ = n × Q ₁ (where n is a natural number), the image quality of re-encoding is greatly improved by determining the quantization step size Q ₂ . be able to.

[0010] quantization step size Q ₁ in the front stage coding may be estimated from the properties of the input image information, it may be extracted what is supplied with image information to be input.

[0011]

FIG. 1 is a block diagram schematically showing an image information re-encoding device according to an embodiment of the present invention.

In FIG. 1, reference numeral 10 denotes image information having no coding history (original image information) or image information having an MPEG-2 coding history accompanied by a preceding coding parameter or accompanied by a preceding coding parameter. The input line 11, which is input without any problem, extracts the encoding parameters in the pre-encoding inputted together with the image information for each frame, or estimates the encoding parameters in the pre-encoding from the properties of the inputted image information. The pre-processing unit 12 receives image information having no encoding history or image information having an encoding history, and an encoding unit that actually performs encoding for information compression. An encoding control unit which controls the encoding operation of the encoding unit 12 based on the encoding parameters and the like at the preceding stage. Ream respectively show output lines is output. When image information (original image information) having no encoding history is input, the apparatus of FIG. 1 functions as a first encoding apparatus, and when image information having an encoding history is input, the apparatus of FIG. The device acts as a re-encoder.

FIG. 2 shows a pre-processing unit 11 according to this embodiment.
FIG. 3 is a block diagram schematically showing an example of the configuration of FIG.

In FIG. 1, reference numeral 20 denotes a detection / classifier for detecting and classifying whether or not input image information is accompanied by coding parameter information of a preceding stage. It is estimated whether the image is processed by the picture type (I / P / B)
Picture type estimator 21b for outputting P / B information, 21b extracts picture type (I / P / B) information of each frame in the preceding coding as a preceding coding parameter and outputs the I / P / B information Picture type reader,
22a is a quantization step size estimator for estimating a quantization step size (Q) in pre-stage encoding as a preceding-stage encoding parameter and outputting Q information, and 22b is a quantization step size of the preceding stage encoding as a preceding-stage encoding parameter. A quantization step size reader 23a for extracting size (Q) information and outputting the Q information, 23a for estimating a block boundary in the previous-stage coding as a preceding-stage coding parameter and outputting image block boundary information Bowl, 23
“b” denotes a block boundary readout unit that extracts image block boundary information in the former-stage coding as a preceding-stage coding parameter and outputs the image block boundary information.

Under the control of the detection classifier 20, when the input image information does not accompany the preceding coding parameter information (when the coding history exists but does not accompany the preceding coding parameter information, or when the coding history If there is no)
, The picture type estimator 21a, the quantization step size estimator 22a, and the block boundary estimator 23a operate to estimate each encoding parameter. On the contrary,
When the input image information accompanies the preceding-stage encoding parameter information, the picture type reader 21b, the quantization step size reader 22b, and the block boundary reader 2
3b operates to extract each encoding parameter. The picture type reader 21b, the quantization step size reader 22b, and the block boundary reader 23b are configured to simply read the encoding parameters recorded in the header of each frame. The preprocessing unit 11 includes only the picture type estimator 21a, the quantization step size estimator 22a, and the block boundary estimator 23a without providing the picture type reader 21b, the quantization step size reader 22b, and the block boundary reader 23b. May be provided. In this case, encoding parameters are estimated for all input image information.

FIG. 3 is a block diagram schematically showing a configuration example of the picture type estimator 21a.

In this example, picture type estimator 2
1a is for identifying an I picture in intra-frame encoding, and encodes all frames of the input image information in the frame (however, the quantization step size and the like are fixed). And an SNR calculator 31 for calculating an SNR value from an encoded bit stream, and an I picture detector 32 for detecting a phase (frame position) of an I picture from the calculated SNR value.

FIG. 4 shows the characteristics of the SNR value in the intra-frame coding when the I picture is identified. The horizontal axis shows the frame number, and the vertical axis shows the SNR value. The quantization step size, quantization matrix, and image type are used. As can be seen from the figure, the SNR in intra-frame encoding
The value does not depend on the quantization matrix, the pre-stage quantization, the image type, etc., and the value of the frame processed by the I picture at the time of the pre-stage encoding is higher than the frame processed by other pictures (P picture, B picture) have. Therefore, the frame position of the I picture can be easily detected.

The picture type estimator 21a shown in FIG.
Performs only the identification of the I picture, but after the identification of the I picture, the identification of the P picture may be further performed in the same manner. Needless to say, if the I picture and the P picture are identified, the position of the B picture is automatically identified.

FIG. 5 shows a quantization step size estimator 22.
FIG. 3 is a block diagram schematically illustrating a configuration example of FIG.

This quantization step size estimator 22a
Is, for example, DCT processing for input image information,
A pre-processor 50 for performing pre-processing such as ME processing, a signal distribution calculator 51 for calculating a distribution of pre-processed signals, and determining a quantization step size at the time of pre-encoding from the calculated distribution. And a Q determiner 52.

FIG. 6 is a block diagram schematically showing a configuration example of the block boundary estimator 23a.

The block boundary estimator 23a includes a cepstrum calculator 60 for calculating a cepstrum of image information.
And a detector 61 that determines a block distortion position from a cepstrum obtained by calculation and detects an image block boundary. Since the block distortion generated in block coding of an image has periodicity, A block boundary is estimated by obtaining a cepstrum of information. For such cepstrum characteristics of image information, see Oda et al., “Basic characteristics of cepstrum information for image signals”, IEICE General Conference, 1995, D.
-361, p. 87, March 1995.

FIG. 7 is a block diagram schematically showing a configuration example of the encoding control unit 13 in the present embodiment.

The encoding control unit 13 receives the image feature amount and the encoding state from the encoding unit 12, and also receives the encoding parameters in the preceding stage encoding from the preprocessing unit 11, and refers to these as described later. Thus, an appropriate parameter in the encoding of the encoding unit 12 is determined, and the operation of the encoding unit 12 is controlled. Further, when image information having no encoding history is input, the encoding control unit 13 refers to only the control defined by the initial parameters, that is, only the property (image feature amount) and the encoding state of the input image. Then, the same coding control as that of the conventional single coding is performed.

As shown in FIG. 7, the coding control unit 13
Picture for determining the picture type of the current re-encoding for each frame based on I / P / B information in the pre-encoding given from the picture type estimator 21a or the picture type reader 21b of the pre-processing unit 11. Q information in the pre-stage encoding provided from the type determination unit 70 and the quantization step size estimator 22a or the quantization step size reader 22b of the pre-processing unit 11, the image feature quantity and the encoding state provided from the encoding unit 12 A quantization step size determination unit 71 that determines a quantization step size of the current re-encoding for each macro block based on
Based on the image block boundary information in the pre-encoding given from the block boundary estimator 23a or the block boundary readout unit 23b of the pre-processing unit 11, every time a series of coded images is processed, the image block position of re-encoding And an image block position determining unit 72 for determining the image block position.

In accordance with the input I / P / B information, if the pre-stage encoding for the frame is the I-picture type, the picture type determining unit 70 performs the encoding at the time of re-encoding by the I-picture type. Choose to do. As described above, it is most important that the phase of the I picture is the same as that in the case of the preceding coding. However, the P picture and the B picture are also made to have the same phase as that of the preceding coding. Is also good.

As is well known, in MPEG-2 encoding, there are three different types of prediction, namely intra-frame encoding (I), forward inter-frame prediction encoding (P), and bidirectional inter-frame prediction encoding (B). The coding efficiency is improved by periodically combining the types. Video frames predicted and encoded in each picture type have different signal properties from each other.
It is important to select a prediction picture type in consideration of this. That is, the cycle of the picture type, for example, GOP
If the phase and the cycle (N) and the I / P cycle (M) are controlled so as to coincide with the case of the preceding-stage encoding, it is possible to prevent a significant deterioration in image quality.

FIG. 8 shows a comparison of the coded picture quality characteristics at different phases when re-encoding is performed on two types of images with the picture type cycle fixed (N = 15, M = 3). The horizontal axis is the frame number,
The vertical axis shows the SNR values for the two types of images, respectively, and uses the frame offset amount as a parameter. However, the quantization step size is fixed (Q = 12).

As can be seen from the figure, the image quality deterioration is alleviated by matching the phases of the I / P pictures, but in order to minimize the deterioration, the GOP phase (the position of the I picture)
It is also necessary to match.

The quantization step size determination unit 71 selects an optimal quantization step size suitable for a bit rate at the time of re-encoding from the image feature amount and the encoding state supplied from the encoding unit 12. in that case in the present embodiment, the quantization step size Q ₂ at the time of re-encoding, on the quantized step size to Q ₁ at the previous stage coding is determined so as to satisfy the following Q rules. That is, Q ₂ ≧ Q ₁ and Q ₂
= N × Q ₁ (where n is a natural number). In an image having an encoding history, the distribution of signal levels is significantly different from an image having no encoding history due to quantization, so that a monotonous relationship is not established between the quantization step size and the quantization distortion. That is, the coding distortion changes in a complicated manner represented by a certain relational expression according to the combination of the quantization step size at the previous stage and at the time of re-encoding. Therefore, it is necessary to set the quantization step size in consideration of the quantization step size of the pre-encoding as described above.

FIG. 9 shows image quality characteristics when re-encoding using a quantization step size according to the Q rule and when re-encoding using a quantization step size not according to the Q rule. FIG. 4 is a characteristic diagram showing a graph, in which a horizontal axis indicates a bit rate and a vertical axis indicates an SNR value.

As can be seen from the figure, when the re-encoding is performed using the above-described Q rule, the re-encoded image quality at a rate slightly lower than the preceding encoding rate is improved. The grounds for reducing the encoding distortion when re-encoding using the Q rule will be described below.

The distortion in the _first encoding is E ₁ , the quantization step size is Q ₁ , the distortion in the _second encoding is E ₂ , and the quantization step size is Q ₂ .

(Equation 1)

When this equation is represented by a graph, a characteristic diagram of the re-encoding distortion E _{2 with} respect to the quantization step sizes Q ₁ and Q _{2 as} shown in FIG. 10 is obtained. From the figure, Q ₂ is Q
It can be seen that re-coding distortion E ₂ is relatively small when a multiple of _1.

The image block position determining section 72 sets the image block position for re-encoding so that the image block boundary in the preceding encoding is maintained as it is.

The encoding section 12 takes in image information for re-encoding, performs re-encoding using the encoding parameters given from the encoding control section 13, and outputs a re-encoded bit stream. . Other configurations and operations of the encoding unit 12 and the encoding control unit 13 are the same as those of a general encoding device.

The configuration of a general encoding apparatus will be briefly described below. The encoding apparatus includes a prediction unit with motion compensation, an orthogonal transformation unit, a quantization unit, an encoding unit, and a buffer unit. Are connected in series, and the output of the buffer means is fed back to the quantization control means to control the above-mentioned quantization means. The prediction means with motion compensation calculates the M of the input current image and the immediately preceding image.
A motion in units of × M blocks is detected by, for example, a block matching method, a predicted image of the current image is created from the previous image in consideration of the motion, and a difference image between the current image and the predicted image is output. The orthogonal transform unit divides the input difference image into N × N blocks, orthogonally transforms the image for each block by, for example, discrete cosine transform (DCT), and outputs image block information to the quantization unit. The quantization means quantizes the image block information based on the given quantization step size, and outputs the quantized image information. Encoding means,
The quantized image information is subjected to variable-length coding by a Huffman coding method or the like in which the number of continuous zero data and the level of the subsequent non-zero data are combined, and the coded image information is output to the buffer means. The buffer means is constituted by, for example, a first-in first-out (FIFO) memory, temporarily stores the encoded image information, and outputs the image information at a constant bit rate according to the FIFO rule. The quantization control means monitors the occupancy of the buffer means at regular intervals, determines the quantization step size to be given to the quantization means according to the occupancy, and controls the code generation amount.

In the embodiment described above, the picture type, the quantization step size, and the image block boundary are used as the encoding parameters. However, in addition, a picture format, a motion vector, and the like may be used as the encoding parameters. Good.

The embodiments described above all show the present invention by way of example and not by way of limitation, and the present invention can be embodied in other various modifications and alterations. Therefore, the scope of the present invention is defined only by the appended claims and their equivalents.

[0041]

According to the present invention described above in detail, according to the present invention, the quantization step size Q ₂ at the time of re-encoding the quantized step size Q ₁ in the front stage coding, Q ₂ ≧
Q ₁ and Q ₂ = n × Q ₁ (where, n is a natural number) since the decided quantization step size Q ₂ using the rules to be, the quality of the re-encoding can be greatly improved.

[Brief description of the drawings]

FIG. 1 is a block diagram schematically illustrating an image information re-encoding device according to an embodiment of the present invention.

FIG. 2 is a block diagram schematically illustrating a configuration example of a preprocessing unit in the embodiment of FIG. 1;

FIG. 3 is a block diagram schematically illustrating a configuration example of a picture type estimator in a preprocessing unit in FIG. 2;

FIG. 4 is a characteristic diagram illustrating characteristics of an SNR value in intra-frame encoding when an I picture is identified.

FIG. 5 is a block diagram schematically illustrating a configuration example of a quantization step size estimator in the preprocessing unit in FIG. 2;

FIG. 6 is a block diagram schematically illustrating a configuration example of a block boundary estimator in the preprocessing unit in FIG. 2;

FIG. 7 is a block diagram schematically illustrating a configuration example of an encoding control unit in the embodiment of FIG. 1;

FIG. 8 is a characteristic diagram showing a comparison between coded image quality characteristics at different phases when re-encoding is performed with a fixed picture type cycle.

FIG. 9 is a characteristic diagram illustrating image quality characteristics when re-encoding using a quantization step size according to the Q rule and when re-encoding using a quantization step size not according to the Q rule;

FIG. 10 is a characteristic diagram showing characteristics of re-encoding distortion E _{2 with} respect to quantization step sizes Q ₁ and Q ₂ .

[Explanation of symbols]

 DESCRIPTION OF SYMBOLS 10 Input line 11 Preprocessing part 12 Encoding part 13 Encoding control part 14 Output line 20 Detection classifier 21a Picture type estimator 21b Picture type reader 22a Quantization step size estimator 22b Quantization step size reader 23a Block boundary Estimator 23b Block boundary reader

Claims (6)

[Claims] 1. A method for re-encoding image information with encoded history, the quantization step size Q ₂ at the time of re-encoding the quantized step size Q ₁ in the front stage coding, Q ₂ ≧ Q ₁ and Q ₂ = n × Q ₁ (where n is a natural number), and re-encode the input image information using the determined quantization step size Q _2. A method for re-encoding image information as a feature. 2. The method according to claim 1, wherein the quantization step size Q ₁ in the pre-encoding is estimated from the properties of the input image information. 3. The method according to claim 1, wherein the quantization step size Q ₁ in the pre-encoding is supplied together with the input image information. 4. A device for re-encoding image information with encoded history, the quantization step size Q ₂ at the time of re-encoding the quantized step size Q ₁ in the front stage coding, Q ₂ Means for determining so as to satisfy ≧ Q ₁ and Q ₂ = n × Q ₁ (where n is a natural number), and re-input of image information input using the quantization step size Q ₂ determined by the determining means. A re-encoding device for image information, comprising: re-encoding means for performing encoding. 5. The apparatus according to claim 4, further comprising: means for estimating a quantization step size Q ₁ in the preceding-stage coding from properties of input image information. 6. The apparatus according to claim 4, further comprising means for extracting a quantization step size Q ₁ in the pre-encoding supplied together with the input image information.