JP3287836B2 - Image data encoding device - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION The present invention will be described in the following order.

A Field of application in industry B Outline of the invention C Conventional technology (Figs. 10 to 12) D Problems to be solved by the invention (Fig. 11) E Means for solving the problems (Fig. 1) F Function (FIG. 1) G Example (G1) Overall Configuration of Image Information Transmission System (FIGS. 1 to 5) (G2) Quantization Step Size Determination Processing (FIGS. 1 and 6)
(FIG. 7, FIG. 7, FIG. 8 and FIG. 9) (G3) Other Embodiments H Effect of the Invention A Industrial Field of the Invention The present invention relates to an image data encoding device, and particularly to a video signal encoding device which encodes a video signal with high efficiency. This is suitable for application when performing conversion processing to image data.

B Overview of the Invention Obtaining evaluation data indicating how much data amount is obtained as encoded data when encoding the input image data,
A quantization step for the input image data is determined using the evaluation data.

C Conventional Techniques Conventionally, in videophone systems and conference telephone systems, relatively strict restrictions have been placed on transmission capacity by efficiently encoding video signals consisting of moving images into intra-frame encoded data and inter-frame encoded data. A video signal transmission system for transmitting a moving image video signal through a certain transmission path has been proposed (Japanese Patent Application Laid-Open No. 63-1183).

That is, for example, as shown in FIG.
_{= T 1, t 2, t} 3 each image PC1 constituting the video at ......, P
When transmitting C2, PC3, etc., the transmission efficiency is improved by compressing the image data to be transmitted using the point that the video signal has a characteristic that the autocorrelation is large as time passes. The intra-frame encoding process performs a compression process on the images PC1, PC2, PC3,..., For example, by comparing pixel data with a predetermined reference value to obtain a difference. , PC2, PC3,... Are transmitted using compressed autocorrelation between pixel data in the same frame.

In addition, the inter-frame encoding process obtains image data PC12, PC23,..., Which are differences in pixel data between sequentially adjacent images PC1, PC2, PC2, and PC3, as shown in FIG. This for the initial image PC1 at the time t = t ₁ transmitting with intraframe coded processed image data.

Thus, the images PC1, PC2, PC3,... Can be highly efficiently coded into digital data having a significantly smaller data amount than when all the image data is transmitted, and transmitted to the transmission path.

Such an encoding process of the video signal is executed in the image data generating device 1 having the configuration shown in FIG.

The image data generator 1 processes the input video signal VD in the pre-processing circuit 2 to perform one-field dropping processing and one-field line thinning processing, and then converts the luminance signal and the chroma signal into 16 pixels (in the horizontal direction). The data is converted into transmission unit block data (referred to as macroblock) data S11 composed of data of × 16 pixels (in the vertical direction) and supplied to the image data encoding circuit 3.

The image data encoding circuit 3 receives the predicted current frame data S12 formed in the prediction encoding circuit 4 and calculates the difference from the macroblock data S11 to generate inter-frame encoded data (this is inter-frame encoded data). (Referred to as an encoding mode) or the difference between the macroblock data S11 and the reference value data to form intra-frame encoded data, which is supplied to the transform encoding circuit 5 as difference data S13.

The transform coding circuit 5 is composed of a discrete cosine transform circuit. The transform coding data S14 obtained by performing high-efficiency coding by orthogonally transforming the difference data S13 is supplied to the quantizing circuit 6 so as to obtain quantized image data. Send S15.

Thus, the quantized image data obtained from the quantization circuit 6
S15 is again subjected to high-efficiency encoding processing in the re-transform encoding circuit 7 including the variable-length encoding circuit, and then supplied to the transmission buffer memory 8 as transmission image data S16.

In addition, the quantized image data S15 is subjected to inverse quantization and inverse transform coding in the predictive coding circuit 4 to be decoded into differential data, and then to correct the unpredicted frame data using the differential data. Thus, new pre-prediction frame data is stored, and prediction is performed by motion-compensating the pre-prediction frame data stored in the prediction encoding circuit 4 with motion detection data formed based on the macroblock data S11. The current frame data can be formed and supplied to the image data encoding circuit 3, whereby the difference between the macroblock data S11 of the frame to be currently transmitted (that is, the current frame) and the predicted current frame data S12 is calculated as difference data. It is made to get as S13.

11 in the configuration of Figure, when transmitting the moving image as described above for FIG. 10, when the first image data of the image PC1 at the time t ₁ of FIG. 10 (A) is given as a macro Bro poke data S11, the image data The encoding circuit 3 enters the intra-frame encoding mode and supplies it to the transform encoding circuit 5 as differential data S13 subjected to intra-frame encoding processing, thereby passing through the quantization circuit 6 and the re-transform encoding circuit 7. Then, the transmission image data S16 is supplied to the transmission buffer memory 8.

At the same time, the quantized image data S15 obtained at the output terminal of the quantizing circuit 6 is subjected to predictive encoding processing in the predictive encoding circuit 4, whereby the unpredicted frame representing the transmitted image data S16 transmitted to the transmission buffer memory 8 is transmitted. data is held before the frame memory, when the macro Bro poke data S11 representing an image PC2 Subsequently at time t ₂ is supplied to the image data coding circuit 3, the predicted current frame data S12
Is supplied to the image data encoding circuit 3 which has been motion compensated.

Thus the image data coding circuit 3 at time t = t ₂
Supplies the differential data S13 subjected to the inter-frame encoding processing to the transform encoding circuit 5, whereby differential data representing a change in the image between the frames is supplied to the transmission buffer memory 8 as transmission image data S16, and When the quantized image data S15 is supplied to the prediction encoding circuit 4, the prediction encoding circuit 4 forms and stores the pre-prediction frame data.

By repeating the same operation thereafter, while the image data encoding circuit 3 executes the inter-frame encoding process, only the difference data representing the change of the image between the previous frame and the current frame is transmitted to the transmission buffer memory 8. Will be sequentially transmitted.

The transmission buffer memory 8 stores the transmission image data S16 sent out in this manner, and sequentially stores the stored transmission image data S16 at a predetermined data transmission speed determined by the transmission capacity of the transmission line 9 as transmission data D _TRANS. And transmit it to the transmission line 9.

At the same time, the transmission buffer memory 8 detects the amount of remaining data, feeds back the remaining data S17 that changes according to the remaining data amount to the quantization circuit 6, and performs quantization step size according to the remaining data S17. By adjusting the amount of data generated as the transmission image data S16, the transmission buffer memory 8 can maintain a proper amount of data (a data amount that does not cause overflow or underflow) in the transmission buffer memory 8. It has been done.

When the remaining amount of data in the transmission buffer memory 8 has increased to the permissible upper limit, the step size of the quantization step STPS (FIG. 12) of the quantization circuit 6 is increased by the remaining amount data S17. The coarse amount of quantization is performed in the quantum circuit 6 to reduce the data amount of the transmission image data S16.

Conversely, when the remaining amount of data in the transmission buffer memory 8 has decreased to the allowable lower limit, the remaining amount data S17 controls the quantization step STPS of the quantization circuit 6 so that the step size becomes smaller. Thus, the amount of data generation of the transmission image data S16 is increased by causing the quantization circuit 6 to execute fine quantization.

D. Problems to be Solved by the Invention As described above, the conventional image data generating apparatus 1 is capable of generating a significant image while matching the data transmission speed of the transmission data D _{TRANS to} the transmission condition limited based on the transmission capacity of the transmission line 9. As a means for transmitting information, the generated image data is stored in the transmission buffer memory 8 so that the image data having a data amount corresponding to the transmission capacity of the transmission line 9 is always extracted by the transmission capacity of the transmission line 9. However, when the signal value represented by the image data to be actually transmitted is large, if the signal value is quantized as it is, the data amount stored in the transmission buffer memory 8 by the large signal value. May be excessively large, and it is desirable to be able to compress.

The present invention has been made in view of the above points, and it is possible to generate quantized image data having an appropriate data amount by controlling the quantization step size according to the signal value of the image data to be quantized. Is to try.

E. Means for Solving the Problem In order to solve the problem, according to the present invention, input image data is converted into encoded image data by a transmission medium 43.
Image data encoding device for transmitting to decoding device 21 via
At 21, an arithmetic means 30 for calculating evaluation data MBP for estimating the amount of data as encoded data when the input image data is encoded from pixel data of the input image data, and a designated quantization step STPS Quantizing means 37 for quantizing input image data on the basis of the above, encoding means 38 for encoding the quantized image data quantized by the quantizing means 37, and encoding by the encoding means 38. Decoding apparatus for converting encoded image data through a transmission medium 43
Buffer means 32 for temporarily storing the encoded image data for transmission to 21, and evaluation data MBP and buffer means 3
When the evaluation data MBP indicates that the input image data has a relatively large data amount based on the buffer remaining amount of 2, the quantization means 37 has a relatively large quantization step. Is specified, or indicates that the evaluation data MBP does not become a large data amount when encoding the input image data, a relatively small quantization step STPS is specified for the quantization means 37. Control means 36
Is provided.

F function Evaluation data MBP for estimating the amount of data as encoded data when the input image data is encoded is obtained, and the quantization step STPS for the input image data is determined using the evaluation data MBP. Thus, the coded image data having an appropriate data amount can always be supplied to the buffer means 32.

G Example Hereinafter, an example in which the present invention is applied to a videophone will be described in detail with reference to the drawings.

(G1) Overall Configuration of Image Information Transmission System In FIGS. 1 and 2, the image information transmission system 21 is composed of an encoder 21A and a decoder 21B, and the encoder 21A converts an input video signal VD _IN into an input circuit unit 22. After the pre-processing, the analog / digital conversion circuit 23 sends the transmission unit block data composed of pixel data of 16 × 16 pixels, that is, the input image data S21 composed of the pixel data of the macroblock MB, to the pixel data processing system SYM1, At each processing stage of the pixel data processing system SYM1, processing information data corresponding to the data to be processed is sequentially transmitted via the header data processing system SYM2 at a timing when the pixel data is processed in units of macroblocks MB. Thus, the pixel data and the header data are respectively transferred to the pixel data processing system SYM1 and the header data. Processing is performed in the data processing system SYM2 according to the pipeline method.

In the case of this embodiment, the macroblock data sequentially transmitted as the input image data S21 is extracted from the frame image data FRM by a method as shown in FIG.

First, one frame image data FRM is divided into two (in the horizontal direction) × 6 (in the vertical direction) block groups GOB as shown in FIG.
OB is 11 (horizontally) x 3 as shown in Fig. 3 (B).
(Vertically) macroblocks MB, each macroblock MB having 16 blocks as shown in FIG. 3 (C).
× 16 pixels of luminance signal data Y _{00 to} Y ₁₁ (8 × 8
And luminance signal data Y).
₀₀ to Y comprising the color signal data C _b and C _r consisting of the corresponding color signal data to all the pixel data of _11.

Thus, the input image data S21 transmitted for each macroblock MB is given to the motion compensation circuit 25,
5 is a motion detection control signal given from a motion compensation control unit 26 provided for the header data processing system SYM2.
In response to S22, the pre-prediction frame data S23 of the pre-prediction frame memory 27 is compared with the input image data S21 to detect the motion vector data MVD (x) and MVD (y), and to the motion compensation control unit 26 for the first. Is supplied as data of the header data HD1 (FIG. 4), and the motion compensation circuit body 25A performs motion compensation for the motion vector data MVD (x) and MVD (y) on the frame data S23 before prediction. The current frame data S24 is formed and supplied to the image data encoding circuit 28 together with the current frame data S25 composed of the input image data S21 to be currently processed.

Here, as shown in FIG. 4, the motion compensation control unit 26 includes, for each macroblock currently being processed as the first header data HD1, the transmission frame number data TR Counter indicating the transmission order of the frame image data FRM, By adding the block group number data GOB address representing the block group GOB (FIG. 3 (A)) and the macro block number data MB address representing the macro block MB among them, the pixel data processing system SYM1 is sequentially processed. The macro block MB transmitted to each processing stage is displayed, and the macro block to be processed is displayed.
Flag data FLAGS indicating MB processing or processing format,
Motion vector data MVD (x) of the macro block MB
And MVD (y) and difference data Σ | AB
| And form

The flag data FLAGS has a maximum of 1 as shown in FIG.
A flag for a word (16 bits) can be set, and a motion compensation control flag MC on / off indicating whether or not the processing target macroblock MB should be processed in the motion compensation mode is set in the 0th bit. You.

The first bit of the flag data FLAGS contains an interframe / intraframe indicating whether the processing target macroblock MB should be processed in the interframe coding mode or the intraframe coding mode. Flag Inter /
Intra is set.

In the second bit of the flag data FLAGS, a filter flag Filter on / off indicating whether or not to use the loop filter 25B of the motion compensation circuit 25 is set.

The third bit of the flag data FLAGS also blow included in the process target macroblock poke data Y ₀₀ -C _r
A transmission flag Coded / Not-coded indicating whether or not (FIG. 3 (C)) should be transmitted can be set.

In the fourth bit of the flag data FLAGS, a drop frame flag indicating whether or not the target macroblock MB is dropped can be set.

Further, in the fifth bit of the flag data FLAGS, a forced refresh flag Refresh on / off indicating whether or not to perform a forced refresh of the macroblock MB to be processed can be set.

Further, a macro block power evaluation flag MBP appreciate can be set in the sixth bit of the flag data FLAGS.

The difference data Σ | A−B | is the difference between the macroblock data A to be currently processed in the current frame data S25 and the macroblock data B compensated by the motion vector for detection in the frame data S23 before prediction. , Which can be used to evaluate the detected motion vector.

The image data encoding circuit 28 outputs the current frame data S2 given from the motion compensation circuit 25 in the intra-frame encoding mode.
5 is supplied as it is to the transform coding circuit 29 as difference data S26, and in contrast, in the inter-frame coding mode, the pixel data of the current frame data S25 and the predicted current frame data S
The difference data S 26, which is a difference from the 24 pixel data, is supplied to the transform encoding circuit 29.

The image data encoding circuit 28 is included in the header data processing system SYM2.
An inter-frame / intra-frame encoding control unit 30 is provided so as to correspond to the image data. Based on the header data HD1 supplied from the motion compensation control unit 26 and the arithmetic data S31 supplied from the image data encoding circuit 28, the An inter / intra flag Inter / Intra (FIG. 5) for designating the encoding mode of the encoding circuit 28 and a filter flag Filter on / off (FIG. 5) for controlling the operation of the loop filter 25B of the motion compensation circuit 25. 5) is calculated and sent to the filter control unit 31 as second header data HD2.

As shown in FIG. 4, the second header data HD2 is composed of the transmission frame number data TR Co constituting the header data HD1.
power data necessary for the filter control unit 31 to generate the inter-frame / intra-frame coding mode switching signal S33 and the filter on / off signal S34, while taking over the unter ~ difference data {| A-B |
(A) ² (L) and Σ (A) ² (H), Σ (AB) ²
(L) and Σ (AB) ² (H), Σ (A-FB)
² (L) and Σ (A-FB) ² (H), Σ (A) are added in the inter-frame / intra-frame coding control unit 30.

Here, the power data Σ (A) ² (L) and Σ (A)
² (H) indicates the lower bit and the upper bit of the square sum of the macroblock pixel data A of the current frame data S25,
Power data Σ (AB) ² (L) and Σ (AB) ²
(H) shows the lower bits of the sum of squares of the difference AB between the macroblock pixel data A of the current frame data S25 and the macroblock pixel data B of the predicted current frame data S24 formed without passing through the loop filter 25B; Represents the high-order bits and represents power data Σ (A-FB) ² (L) and Σ (AF)
B) ² (H) is the macroblock pixel data A of the current frame data S25 and the macroblock pixel data FB of the predicted current frame data S24 formed via the loop filter 25B.
And the power data Σ (A) represents the sum of the macroblock pixel data A of the current frame data S25. These power data express the data amount as a power value in order to evaluate the size of the data to be processed (the sum of squares was obtained as a value irrelevant to the sign), and the input image data was encoded. At this time, it is used as evaluation data indicating the amount of data as encoded data.

The filter control unit 31 receives the second header data passed from the inter-frame / intra-frame coding control unit 30.
On the basis of HD2 and the remaining amount data S32 supplied from the transmission buffer memory 32, an inter-frame / intra-frame encoding mode switching signal S33 is sent to the image data encoding circuit 28, and the loop filter 25B is sent to the loop filter 25B. A filter on / off signal S34 is transmitted, and a filter flag Filter on / o indicating the content of the filter on / off signal S34 is provided.
ff is added to the second header data HD2 and passed to the threshold control unit 35 as the third header data HD3.

Here, the filter control unit 31 first activates the image data encoding circuit 28 when the amount of transmission data in the case of performing inter-frame encoding processing is larger than the amount of transmission data in the case of performing intra-frame encoding processing. Control to intra-frame encoding mode.

Secondly, the filter control unit 31 is configured such that in the state where the processing is performed in the inter-frame coding mode, the predicted current frame data S24 that has not undergone the processing is better than the predicted current frame data S24 that has been processed in the loop filter 25B. If the difference value is small, the filter on / off signal S34
The loop filter 25B is controlled so as not to perform the filtering operation.

Third, when the filter control unit 31 enters the forced refresh mode, the image data encoding circuit 28 is switched by the inter-frame / intra-frame encoding mode switching signal S33.
To the intra-frame encoding mode.

Fourth, when the transmission buffer memory 32 has a possibility of overflowing based on the remaining amount data S32 supplied from the transmission buffer memory 32, the filter control unit 31 should detect this and perform frame drop processing. Third header data comprising a flag for instructing
HD3 is sent to the threshold control unit 35.

Thus, the image data encoding circuit 28
The difference data S26 encoded in a mode that minimizes the difference between S25 and the predicted current frame data S24 is supplied to the transform encoding circuit 29.

As shown in FIG. 4, the third header data HD3 is obtained from the transmission frame number data TR Counter from the header data HD2.
Together take over the ~ motion vector data MVD (x) and MVD (y), the filter flag Fil of 6 bits fraction in the filter control Yunitsuto 31 corresponding to Bro poke data Y ₀₀ -C _r
ter on / off is added.

Transform coding circuit 29 is made of a discrete cosine transform circuit, block the coefficient value after the discrete cosine transform of the six _{Y 00, Y 01, Y 10} , Y 11, C b, and a zigzag-scan for each C _r It is transmitted to the transmission block setting circuit 34 as transformed encoded data S35.

Transmission block setting circuit 34 six Bro poke is coming sent as converted coded data S35 data Y ₀₀ -C _r (3
(C)), from the first coefficient data to n
The sum of the squares is calculated, and the result of the calculation is passed to the threshold control unit 35 as power detection data S36.

At this time Sureshiyorudo control Yunitsuto 35 compares the power detection data S36 to a predetermined Sureshiyorudo each blow poke data Y ₀₀ -C _r, do not allow transmission of the blow stick data when power detection data S36 is smaller than the Sureshiyorudo, which The transmission permission / prohibition data CBPN of 6 bits indicating that the transmission is permitted when it is large is added to the third header data HD3 passed from the filter control unit 31, and the quantization control is performed as the fourth header data HD4. together passed to Yunitsuto 36, thereby delivering the blow poke data Y ₀₀ -C _r corresponding from the transmission block setting circuit 34 as transmission Bro poke patterned data S37 to the quantization circuit 37.

Here, the fourth header data HD4 is the transmission frame number data TR Coun of the header data HD3 as shown in FIG.
ter ~ Inherit the filter flag Filter on / off as it is, and block the threshold control unit 35.
Y ₀₀ -C _r the transmission availability flag CBPN of 6 bits fraction was generated corresponding is added.

Quantization control unit 36 is threshold control unit 35
The quantization step size determination processing routine RT0 shown in FIG. 6 is performed based on the fourth header data HD4 passed from the controller and the remaining amount data S32 transmitted from the transmission buffer memory 32.
Is performed to perform the quantization step size control signal S3.
8 to the quantization circuit 37, thereby causing the quantization circuit 37 to perform quantization processing with a quantization step size adapted to the data included in the macroblock MB, and as a result, the quantization obtained at the output end of the quantization circuit 37. The image data S39 is supplied to the variable length encoding circuit.

At the same time, as shown in FIG. 4, the quantization control unit 36 converts the fifth header data HD5 into the header data HD4.
Bro poke data Y ₀₀ -C _r (FIG. 3 (C)) to the corresponding flag data FLAGS and the motion vector data MVD (x) and MVD (y) in this separated were arranged in series data on the basis of the Is formed and passed to the variable length coding circuit 38 and the inverse quantization circuit 40.

Here, the header data HD5 is, as shown in FIG.
Transmission frame number data TR Coun of the header data HD4
ter~ macroblock number as with taking over the data MB address, a quantization size data QNT in the quantization control Yunitsuto 36, Bro poke data Y ₀₀ -C flag data FLAGS for _r, the motion vector data MVD (x) and MV
D (y) is added.

The variable-length encoding circuit 38 subjects the header data HD5 and the quantized image data S39 to variable-length encoding and performs transmission image data S40.
And supplies it to the transmission buffer memory 32.

Variable-length coding circuit 38 during the variable-length coding the blow poke data Y ₀₀ -C _r, the corresponding flag data "frame dropping" on the basis of the FLAGS, or when the "not sent" is specified, the block The data is discarded without being transmitted as the transmission image data S40.

The transmission buffer memory 32 stores the transmission image data S40, reads out the transmission image data S40 at a predetermined transmission speed, combines the transmission image data S40 with the transmission audio data S41 transmitted from the audio data generator 42 in the multiplexer 41, and transmits the resultant data to the transmission line 43.

The inverse quantization circuit 40 inversely quantizes the quantized image data S39 sent from the quantization circuit 37 based on the header data HD5, and then supplies the inversely quantized data S42 to the inverse transform encoding circuit 43 to perform inverse quantization. After being converted into the converted coded data S43, it is supplied to the decoder circuit 44, and thus the coded difference data S44 representing the image information transmitted as the transmission image data S40 is supplied to the pre-prediction frame memory 27.

At this time, the pre-prediction frame memory 27 uses the coded difference data S44 to correct the pre-prediction frame data stored up to that time, and stores it as new pre-prediction frame data.

Thus, according to the encoder 21A having the configuration shown in FIG. 1, pixel data is pipelined in macroblock units in the pixel data processing system SYM1 based on the header information supplied from the header data processing system SYM2. By passing the header data in the header data processing system SYM2 so as to synchronize with this, the pixel data is added or deleted as necessary by adding or deleting the header data as necessary in each processing stage of the header data processing system SYM2. Adaptive processing is possible.

As shown in FIG. 2, the decoder 21B receives the transmission data transmitted from the encoder 21A via the transmission line 43 into the transmission buffer memory 52 via the demultiplexer 51, and also transmits the transmission audio data S51 to the audio data receiving device 53. To receive.

The image data received by the transmission buffer memory 52 is separated into received image data S52 and header data HD11 by a variable length inverse transform circuit 54, and is inversely quantized into inverse quantized data S53 by an inverse quantization circuit 55, and then an inverse transform coding circuit. Discrete inverse transform processing at 56 and inverse transform encoded data
It is converted back to S54.

The inverse transform encoded data S54 is supplied to the decoder circuit 57 together with the header data HD12 formed in the inverse quantization circuit 55, and is encoded as encoded differential data S55 by the frame memory 5.
Stored in 8.

Thus, the encoded difference data S55 is stored in the frame memory 58.
The decoded image data S56 is converted into an analog signal in a digital / analog conversion circuit 59, and then output from the output circuit section 60.
Is transmitted as an output video signal VD _OUT via the.

(G2) Quantization step size determination processing The quantization control unit 36 has the macro block MB currently being processed by executing the quantization step size determination processing routine RT0 shown in FIG. 6 for each macro block MB. By selecting a quantization step size QNT suitable for the format of the image data (this is called a macro block type) and supplying it to the quantization circuit 37 as a quantization step size control signal S38, the macro block type is selected. Then, the quantization circuit 37 is controlled so as not to cause image quality disturbance that may occur.

In this embodiment, as shown in FIG. 7, the quantization circuit 37 can vary the steps from the upper limit value QNT = 31 to the lower limit value QNT = 1 as the quantization step size QNT.
The quantization control unit 36 sets the data remaining amount Buffer of the transmission buffer memory 32 to a value corresponding to the variable control range of the quantization step size QNT, that is, an appropriate value that falls within the range of the quantization size controllable range QCR. Macro block type Quantization step size QNT value
Adaptive control is performed according to e.

(G2-1) Processing when the remaining data amount is excessive, that is, when the quantization control unit 36 enters the quantization step size determination processing routine RT0 in FIG. 6, it proceeds to step SP1.
Then, it is determined whether or not the remaining amount data Buffer of the transmission buffer memory 32 is larger than the sum of the margin Margine and the quantization size controllable range QCR.

If a positive result is obtained here, this means that the remaining amount of data Buffer in the transmission buffer memory 32 has exceeded the upper limit value.
After moving to SP2, a quantization step size control signal S38 for setting the quantization step size QNT to the maximum value, that is, QNT = 31, is supplied to the quantization circuit 37, and then, to step SP3, the currently set quantization step is processed. Save the size QNT as the previous frame quantization step size PQNT.

Thus, the quantization control unit 36 ends the quantization step size determination processing routine RT0 in step SP4, and thereby the quantization circuit 37 performs quantization of the transform coded data S35 with the coarsest quantization step size.

As a result, the data amount of the quantized image data S39 generated from the quantization circuit 37 is controlled to the smallest value, so that the remaining data amount Buffer of the transmission buffer memory 32 decreases.

This operation is repeatedly executed while a positive result is obtained in step SP1, and as a result, the remaining data in the transmission buffer memory 32 eventually becomes smaller than the sum QCR + Margine of the margin Margin and the quantization size controllable range QCR.

(G2-2) Processing in Intra-frame Coding Mode In such a state, the quantization control unit 36 moves to step SP5 when a negative result is obtained in step SP1, and the macro block type Macro Block Typ
It is determined whether or not e is an intra-frame coding block and a block which is not a forced refresh block not refresh block.

Here, as shown in FIG. 8, the macro block type is the second bit, first bit, and zero bit of the flag data FLAGS included in the header data HD4 passed from the threshold control unit 35 to the quantization control unit 36. These bits are represented by bits 01
When `` 0 '', the macro block type is intra-frame coding type I
ntra, “000” indicates inter-frame coding type Inter, and “001” indicates filter-free motion compensation type MC-not.
It is filtered, and when it is "101", it becomes a filter-use motion compensation type MC-filtered.

Therefore, when a positive result is obtained in step SP5, this means that the second, first, and 0th data in the flag data FLAGS
This indicates that the bit is “000” and the forced refresh flag refresh of the fourth bit is in the state of logic “0”.

By the way, such a state means that the change of the current frame with respect to the previous frame is so severe that the macroblock type requires intra-frame coding, and moreover, the forced refresh mode is not currently specified. Under the condition.

Under these conditions, the quantization circuit 37 performs quantization with a fine quantization step size.
It can be said that the data amount of the quantized image data S39 generated from 37 becomes an extremely large value, and the possibility that the buffer memory 32 eventually overflows is approaching.

At this time, the quantization control unit 36 proceeds to step SP6 and sets the quantization step size QNT to the upper limit value QNT = 31.
The quantized image data generated by the quantization circuit 37
A process for suppressing the data amount in S39 is executed, and as a result, the occurrence of a state in which the transmission buffer memory 32 overflows is avoided.

On the other hand, when a negative result is obtained in step SP5, this indicates that the type of the macroblock to be processed is not the intra-frame coded Intra, or that the result of the forced refresh even if the intra-frame coded Intra is used. This indicates that this has occurred, and at this time, the quantization control unit 36 jumps this without performing the processing in step SP6.

(G2-3) Processing in forced refresh mode Next, in step SP7, the quantization control unit 36 determines that the type of macro block to be processed is a forced refresh type.
Judge whether it is a fresh block or not.

If a positive result is obtained here, it indicates that it is specified that the forced refresh should be performed.At this time, the quantization control unit 36 proceeds to step SP8 and sets the quantization step size QNT as that of the previous frame. Previous frame quantization step size PQNT used in quantization processing
Is set, and when it is specified that the forced refresh process is to be executed, the quantization is executed with the same quantization step size as the previous frame.

In this way, when the forced refresh process is executed, it is possible to prevent the image quality from being changed so that the forced refresh process does not become a problem in practical use.

However, since the forced refresh is executed at a predetermined period and independently of the image content, if the quantization step size changes compared to the previous frame even though the image content does not change, the relevant refresh is performed. Changes are often annoying.

On the other hand, if the value of the quantization step size is not changed when the forced refresh is designated, it is possible to prevent an unsightly change in the image at the time of the refresh.

Note that the same effect as in the above case can be obtained by selecting a slightly smaller value instead of selecting the same value as the value of the quantization step size.

When the forced refresh is specified in the state where there is no change in the image, if the quantization step size is increased, this will degrade the quality of the restored image, while the quantization step size will be slightly reduced. If the size is reduced, the image quality of the restored image can be slightly improved, so that practically human eyes can not feel the change as the image.

On the other hand, if a negative result is obtained in step SP7, this indicates that the forced refresh is not currently specified, and at this time, the quantization control unit
36 jumps this without performing the processing in step SP8.

(G2-4) Processing when power of difference data is large Next, in step SP9, the quantization control unit 36 determines that the macroblock type is the inter-frame coding type Inter and the macroblock power MBP is equal to the predetermined threshold value T.
Determine if it is greater than hreshold.

Here, the macroblock power MBP is When a positive result is obtained in step SP9, this means that the image data of the macroblock MB, that is, the macroblock power MBP of the difference data is large, so that the quantization is performed with a somewhat coarse quantization step size. Means that there is no effect that the image quality of the restored image is extremely deteriorated even if the transmission is performed.

Therefore, when a macroblock satisfying such conditions is subjected to discrete cosine transform in the transform coding circuit 29, the quantization control unit 36 executes this step.
After confirming at SP9, the process proceeds to step SP10 to set the quantization step size QNT to the coarsest value, that is, the upper limit value 31.

On the other hand, when the macroblock power MBP whose macroblock power MBP is not so large is transmitted, the quantization control unit 36 jumps this so as not to perform the processing in step SP10.

By the way, the macroblock power MBP expressed by the equation (1) is _calculated based on the discrete cosine transform coefficient _Coeff (i) obtained as a result of the discrete cosine transform process in the transform coding circuit 29. The weight of the discrete cosine transform coefficient indicates the strength of the transmission signal obtained by performing the discrete cosine transform, and therefore, it is difficult for the macroblock power MBP to be large. Since the strength of the transmission signal as the signal transmission means is large, even if the transmission signal is slightly compressed and transmitted, the transmission information can be correctly reproduced on the receiving side without being affected by external noise.

Therefore, in such a case, the quantization control unit 36 changes the quantization step size QNT to a large value, thereby compressing the data amount of the quantized image data S39 generated in the quantization circuit 37, thereby transmitting the data to the transmission path 43. Try to reduce the burden.

The discrete cosine transform circuit constituting the transform coding circuit 29 is given by the following equation: , And the inverse cosine transform circuit forming the inverse transform coding circuit 56 is represented by the following equation: Performs the inverse discrete cosine transform. Here, x and y represent the coordinates of the pixel in the macro block (the coordinates (0, 0) at the upper left corner), and u and v represent the coordinates of the coefficients at the time of discrete cosine transform.

When u and v = 0, In other cases, C (u) C (v) = 1 (5)

In actuality, the transformations of the equations (2) and (3) are performed as follows. When X is an image data matrix in a macro block and C is a transformation matrix for discrete cosine transformation, the transform encoding circuit 29 first After obtaining the converted image data matrix XC- ¹ by horizontally converting the matrix X, the converted image data matrix C (X) C- ^{1 is} then subjected to the vertical conversion process again.
Get.

The converted image data matrix C (X) C thus obtained
^-1 is the coefficient _Coeff (1), C, as shown in FIG.
_oeff (2), C _oeff (3)... C _oeff (64) can be represented as a transform coefficient matrix composed of an 8 × 8 matrix, and each coefficient C _oeff (i) (i = 1 to 64) is read out from the transformation matrix in the order of i = 1, 2, 3,... 64 while scanning over time.

Thus, the image data for one macroblock is converted into a conversion coefficient _Coeff (i) (i = 1 to 64) constituting a conversion matrix, and this is supplied to the quantization circuit 37 as transmission data arranged in time series. Will be.

Thus, the transform coefficient data sequence supplied to the quantization circuit 37
_Coeff (1), _Coeff (2) ... _Coeff (64) represents the information to be transmitted and also represents the strength of the signal to be transmitted. As represented by the equation, the transform coefficient data sequence C _oeff (i) (i
= 1, 2,... 64), the squares of the conversion coefficient data from i = 1 to n are cumulatively added so that the effects of the signal to be transmitted are equal in the horizontal and vertical directions. Equation (1) eventually gives this value to the macroblock power
It is defined as MBP.

In practice, when a transform coefficient matrix is obtained as shown in FIG. 9 by performing discrete cosine transform of image data, power is concentrated on the transform coefficient C _oeff (i) in the upper left corner, that is, a lower-order transform coefficient. On the other hand, the conversion coefficient at the lower right corner, that is, the higher-order conversion coefficient does not tend to generate significant information, and thus the transmission data can be compressed by the discrete cosine transform.

Accordingly, when it is confirmed in step SP9 in FIG. 6 that the macro block type is the inter-frame coding type Inter, the macro block power MBP obtained based on the equation (1) is set to a predetermined threshold value Thres.
If it can be confirmed that the value is larger than the hold, it means that the value of the difference data in the macroblock MB is sufficiently large, so that coarse quantization may be performed. Therefore, if the quantization step size QNT is selected as the upper limit value in step SP10 according to such a determination,
The difference data can be transmitted with a relatively small data amount.

(G2-5) Processing in Interframe Coding Mode In step SP11 of FIG. 6, the quantization control unit 36 determines that the macroblock type Macro Block Type is the interframe coding type Inter and the macroblock power MBP is a predetermined threshold. When it is confirmed that the value is smaller than the threshold value, the process proceeds to step SP12, where the quantization step size QN is set.
Set T to the value of 1/2 of the previous frame quantization step size PQNT used in the previous frame, and then
In P13, it is determined whether or not the value of the set quantization step size QNT is smaller than the lower limit value 1. When smaller, the quantization step size QNT is set to the lower limit value 1 in step SP14.
And if the value is not smaller than 1, the quantization step size QNT is used without resetting.
The processing to set as is.

Here, since the macroblock power MBP indicates the strength of the difference data signal of the image data of the macroblock as described above with respect to the equation (1), the difference is small when a positive result is obtained in step SP11. Therefore, the change of the content of the image is smaller than that of the image of the previous frame.

When such image data is obtained, it means that the image to be transmitted at present is in a state in which the image of the previous frame has been changed only partially so as to be modified without largely changing the image of the previous frame. ing.

Then, if the quantization control unit 36 reduces the quantization step size PQNT to 1/2 based on the quantization processing result of the previous frame in step SP12 and sets it as the quantization step size QNT of the current frame, Since the quantization step size can be made smaller for the current frame in which the motion has decreased compared to the previous frame, the quantization step size can be set to an even more optimal quantization step size.

Since such a quantization step size reduction process is repeatedly performed by the quantization control unit 36 in steps SP11 and SP12 as long as an image with little motion continues thereafter,
Eventually, when the image with little motion is continuously transmitted, the quantization step size can be converged to a value suitable for this.

Thus, in step SP13 and step SP14, the quantization step size QNT is not made smaller than the lower limit value 1, so that the quantization control unit 36 eventually turns to the quantization step size when transmitting an image with little motion.
The quantization process can be performed stably with QNT converging to the lower limit.

On the other hand, if a negative result is obtained in step SP11, the quantization control unit 36 determines that the change of the image to be transmitted at present is large, and the steps SP12, SP13, SP
Jump this without doing the processing of 14.

(G2-6) Operation of the quantization control unit 36 In FIG. 6, first, when the remaining data Buffer of the transmission buffer memory 32 exceeds the upper limit (QCF + Margin),
The quantization control unit 36 detects this in step SP1, sets the quantization step size QNT of the quantization circuit 37 to the upper limit value 31 in step SP2, and thereby reduces the remaining data buffer in the transmission buffer memory 32. Is controlled so as to be maintained at or below the upper limit.

In this state, second, the macroblock type Macr
o When block data whose Block Type is an intra-frame encoding type and is not a forced refresh block is given to the quantization circuit 37, the quantization control unit 36
Check at SP5 and at step SP6 quantize circuit 37
By setting the quantization step size QNT to the upper limit value 31, the transmission buffer memory 32 is controlled not to overflow.

In such an operation mode, the quantization control unit 36 obtains a negative result in each of the steps SP7, SP9, and SP11, so that the quantization step size set in the step SP6 is used as the previous frame quantization step size PQNT in the step SP3. After the setting, the processing procedure ends.

Third, when data having a macroblock type of the forced refresh block type is given to the quantization circuit 37, the quantization control unit 36 performs a loop of steps SP1-SP5-SP7 in the quantization step size determination processing routine RT0. In step SP8, the previous frame quantization step size PQNT is set as the quantization step size QNT of the quantization circuit 37, and thereby the image quality of the image transmitted when entering the forced refresh mode is set. By preventing the change from the image of the previous frame, an unsightly change in the image quality during forced refresh is prevented.

At this time, the quantization control unit 36 includes steps SP9 and SP11.
In step SP3, a negative result is obtained, and the quantization step size QNT set in step SP8 is changed to the previous frame quantization step size PQNT in step SP3.
And the process ends.

Fourth, when image data of a macroblock having a large macroblock power MBP is supplied to the quantization circuit 37,
The quantization control unit 36 confirms this by the loop of steps SP1-SP5-SP7-SP9 in the quantization step size determination processing routine RT0, and in step SP10, sets the quantization step size QNT of the quantization circuit 37 to the upper limit value 31. Thus, the amount of data generated in the quantization circuit 37 can be suppressed to a small value, and as a result, image data can be transmitted more efficiently.

At this time, after such processing is completed, the quantization control unit 36 sets the quantization step size QNT set in step SP10 in step SP3 as a previous frame quantization step size PQNT because a negative result is obtained in step SP11. After the correction, the processing routine ends.

Fifth, when macroblock data of a small macroblock power MBP is supplied to the quantization circuit 37 with an interframe coding type, the quantization control unit 36 converts the macroblock data into steps SP1 to SP5 of the quantization step size determination processing routine RT0. −S
Confirm by the loop of P7-SP9-SP11, and step SP12
In (2), the quantization step size is made to converge to the lower limit value 1 by setting a half value of the previous frame quantization step size PQNT as the quantization step size QNT.

Thus, the optimum quantization step size can be set by applying the macroblock power to the MBP.

(G3) Other Embodiments (1) Although the case where the quantization step size QNT is set to the upper limit value 31 in steps SP6 and SP10 in FIG. 6 has been described, the quantization step size QNT to be set is limited to the upper limit value. Instead, other values may be set.
What is necessary is just to select a coarse quantization value large enough to execute the coarse quantization.

(2) In the steps SP9 and SP11 in FIG. 6, the case where the same threshold value Threshold is selected for judging the magnitude of the macroblock power MBP has been described, but a different value may be selected instead. Even in this case, the same effect as in the above case can be obtained.

(3) In step SP12 in FIG. 6, the case where the quantization step size QNT is obtained from the previous frame quantization step size PQNT and a value of 1/2 thereof is set has been described. The quantization step size is not limited to 2 and may be changed to another value as needed. In short, it is sufficient to select a quantization step size having a size reduced by a predetermined ratio with respect to the previous frame quantization step size.

(4) The case where the quantization step size QNT is made to converge to the lower limit value 1 in steps SP13 and SP14 in FIG. 6 has been described. However, the value to be converged is not limited to the lower limit value, and other values may be used as necessary. A selection may be made.

H Advantageous Effects of the Invention As described above, according to the present invention, evaluation data for estimating, from pixel data of input image data, the amount of data to be coded data when encoding the input image data is obtained. By determining the quantization step for the input image data based on the data and the remaining amount of the buffer, it is possible to always supply the coded image data of the appropriate data amount to the buffer means.

[Brief description of the drawings]

1 and 2 are block diagrams showing an encoder and a decoder constituting an image information transmission system to which an image data encoding apparatus according to the present invention is applied, FIG. 3 is a schematic diagram showing a configuration of frame image data, and FIG. 4 is a block diagram showing the header data processing system of FIG. 1, FIG. 5 is a schematic diagram showing the configuration of the flag data of FIG. 4, and FIG. 6 is a quantization step of the quantization control unit 36 of FIG. FIG. 7 is a flow chart showing a size determination processing routine, FIG. 7 is a curve diagram showing a change in the remaining amount data of the transmission buffer memory 32 in FIG. 1, FIG. 8 is a table showing types of macro block type, and FIG. FIG. 10 is a schematic diagram for explaining the intra-frame / inter-frame encoding process, FIG. 11 is a block diagram showing a conventional image data generator, and FIG. 12 is a quantization step thereof. In the curve diagram is there. 21 ... Image information transmission system, 21A ... Encoder, 21B
... A decoder, 25 a motion compensation circuit, 26 a motion compensation control unit, 27 a frame memory before prediction, 28 a picture data encoding circuit, 29 a transform encoding circuit, 30 an interframe / Intra-frame encoding control unit, 31 ... Filter control unit, 32 ... Transmission buffer memory, 34 ... Transmission block setting circuit, 35 ... Threshold control unit, 36 ... Quantization control unit, 37 ... Quantization circuit, 38
.... Variable length coding circuit.

Claims (1)

(57) [Claims] An image data encoding apparatus for transmitting input image data as encoded image data to a decoding apparatus via a transmission medium, wherein the input image data is encoded as encoded data when the encoded image data is encoded. Computing means for computing evaluation data for estimating whether the data amount is equal to the pixel data of the input image data; quantizing means for quantizing the input image data based on a designated quantization step; Encoding means for encoding the quantized image data quantized by the encoding means, and transmitting the encoded image data encoded by the encoding means to a decoding device via a transmission medium. A buffer means for temporarily storing the encoded image data; and the evaluation data being based on the evaluation data and the remaining buffer amount of the buffer means. When it is indicated that a relatively large amount of data is to be obtained when encoding the input image data, the relatively large quantization step is designated for the quantization means, or when the evaluation data is the input image data. Control means for designating the quantization step to be relatively small with respect to the quantization means when it indicates that the data amount will not be large when encoding the image. Data encoding device.