JP4604757B2 - Image coding apparatus and image coding method - Google Patents

Description

  The present invention relates to an image encoding apparatus that encodes image data, and more particularly to an image encoding apparatus that encodes moving image data at a variable bit rate and a method thereof.

Usually, in an image encoding device, an image signal is encoded at a fixed encoding rate (CBR: Constant Bit Rate) or a variable encoding rate (VBR: Variable Bit Rate) and output as a bit stream.
A bit stream encoded at a fixed encoding rate has a constant bit rate and does not deviate significantly from the target bit rate.
On the other hand, since the bit stream encoded at the variable encoding rate changes according to the characteristics of the image signal to be encoded, the bit rate is not constant. However, even in this case, the compression coding process is performed on the image signal within a range not exceeding the maximum bit rate determined by the standard.

When a digitized moving image signal is encoded, encoding bit rate control is performed.
The encoding bit rate control is control to achieve a desired bit rate when a code sequence generated by encoding is viewed in a certain time width.

To date, a plurality of standards for compressing and encoding moving images with high efficiency have been proposed. Among them, the MPEG2 standard is currently attracting attention in various application fields such as communication, broadcasting, and media storage as a standard for moving picture coding, and is being put into practical use.
One application example is the DVD (Digital Versatile Disc), which is a large-capacity, low-priced and easy-to-use recording medium for storing data that is compression-encoded for moving images according to the MPEG2 standard. Is expected to spread.

The MPEG2 standard does not specify encoding control for controlling the bit rate. For this reason, the encoded information amount only needs to be within a range defined by the standard.
Usually, a variable encoding rate is used in DVD authoring (authoring, editing work for creating a DVD title). In this case, the encoding method checks the encoding difficulty level of the entire source to be encoded in the first process, and the entire encoded information amount is within the range defined by the standard in the second process. A variable encoding rate is realized by assigning a code amount corresponding to the above-described encoding difficulty level and performing encoding. By such two processes, a larger amount of code is allocated to a highly complicated (complex) image signal while the entire amount of encoded information is within the range defined by the standard. By suppressing the amount of code assigned to an image signal with low difficulty (uncomplicated), uniform image quality can be achieved as a whole.

  By the way, in an application where a moving image signal obtained by an imaging device such as a video camera is recorded on a recording medium in real time, it is necessary to sequentially compress and encode the moving image signal input continuously and record it on the recording medium. There is. When such real-time processing is required, the first processing in the above-described two-stage encoding processing cannot be performed. In other words, since the entire image to be encoded cannot be examined in advance, the image is usually recorded at a fixed encoding rate in which the code amount is controlled to be constant for each GOP (Group of Picture) in units of about 15 frames. Perform compression encoding processing.

FIG. 1 shows an encoding rate when an image signal is encoded at a fixed encoding rate. When an image signal is compression-encoded by fixed encoding rate control, the bit rate of the obtained code string is almost fixed to the target bit rate regardless of the characteristics of the image. In order to realize this fixed coding rate, the coding rate control unit of the image coding device quickly changes the quantization scale in units of several macroblocks with respect to a predetermined target coding rate. Let me.
The compression encoding process with a fixed encoding rate keeps the encoding rate constant regardless of the image characteristics, so the image quality is good in scenes with low encoding difficulty, but the image quality deteriorates in scenes with high encoding difficulty. This causes a malfunction. For this reason, it is desirable to variably set the bit rate of the encoding process so that the code amount obtained by compression encoding falls within a specified value while maintaining the image quality at a certain level or higher.

FIG. 2 shows an example of the bit rate of a code string obtained by variable bit rate control (variable coding rate control). In the variable bit rate control, the quantization (Q) scale is controlled with respect to the target bit rate, for example, in units of GOP. For example, the bit rate is low where the image encoding difficulty is low (uncomplicated image portion), and the bit rate is high where the image encoding difficulty is high (complex image portion). Since the bit rate changes with time according to the degree of difficulty, the quality of the reproduced image can be kept substantially uniform.
However, in an application example that requires real-time image processing, such as a video camera, as described above, since the characteristics of the entire image cannot be examined in advance, the encoding rate is set empirically. Since the image obtained from the actual imaging device changes from moment to moment, and the image encoding difficulty also changes with time, the encoding bit rate comes to the beginning of the bitstream in the image portion where the encoding difficulty is high The maximum bit rate specified in the sequence header may be exceeded. Therefore, it is impossible to generate a bitstream conforming to the standard in all cases simply by reducing the response speed of the quantization scale with respect to the change in the encoding difficulty level.
Further, considering the case where the variable coding rate control is applied to consumer use, a simple algorithm with a small increase in hardware circuit scale and software calculation amount is required.

  Patent Document 1 discloses an image code that can be used for overcoming the above-described problems and realizing compression coding processing of a moving image with variable coding rate control in an application that requires real-time processing. An apparatus and a processing method thereof are disclosed. In summary, the technique disclosed in Patent Document 1 is a variable bit rate (VBR) control technique performed by controlling a quantization (Q) scale. The outline of the invention disclosed in Patent Document 1 will be described below.

  According to Patent Document 1, an image encoding process using a so-called fixed Q scale that fixes the Q scale to a predetermined value is performed. As a result, more bits are generated in an image with a high degree of difficulty in encoding, and conversely, fewer bits are generated in an image with a low degree of difficulty in encoding. As a result, if you look at the whole series of images, the amount of bits that occur in parts with low encoding difficulty is suppressed, and a larger number of bits are allocated in parts with high difficulty, so it is variable while maintaining uniform image quality. Encoding rate can be realized.

  The technique disclosed in Patent Document 1 performs coding rate control in units of GOPs, for example, units of 15 pictures or more so that the average bit rate of the entire image becomes a constant reference. By setting the target bit amount for each GOP in advance and controlling the Q scale according to the deviation of the generated bit amount of each GOP from the target bit amount, the coding rate changes gradually in GOP units, and the entire image Is maintained within a range that does not deviate significantly from the target bit rate.

  Furthermore, the technique disclosed in Patent Document 1 is such that the second quantization (Q) provided in the normal image coding apparatus so that the coding rate of the input image satisfies the maximum allowable bit rate of the coding system. A fixed coding rate control Q scale is set by the scale setting means in accordance with a predetermined maximum bit rate, and the fixed coding rate control Q scale and the variable code obtained by the first quantization scale control means are set. The bit rate in the image encoding process can be maintained within the maximum allowable value of the encoding system by applying the larger one to the actual encoding process by comparing with the Q scale for encoding rate control. .

In addition, since the variable coding rate processing disclosed in Patent Document 1 uses a normal fixed coding rate control means, when applied to an image coding device, the hardware circuit scale and software computation amount There is an advantage that the increase in the amount can be minimized.
WO00 / 65842 (PCT / JP00 / 02592)

In the technique disclosed in Patent Document 1, the maximum bit rate is limited so as not to exceed the maximum bit rate determined by the standard. However, considering the image quality by encoding, as illustrated in FIG. 3, when an image having a high encoding difficulty level continues and the Q scale is high, an image having a low encoding difficulty level (the Q scale is low). In the first transition region Z1 in which an image that converges to a value) is input, the bit rate becomes too low, and image quality deterioration due to encoding is conspicuous. Therefore, it is necessary to limit the minimum bit rate in order to prevent such image quality deterioration.
On the other hand, if the bit rate is too high in the second transition region Z2 of FIG. 3 even when the target bit rate is low, the generated bit amount in the long term becomes a value corresponding to the target bit rate. Since it is necessary to add a restriction, it is necessary to lower the bit rate thereafter to offset the excessive bit amount, and image quality may be deteriorated at this time.
In other words, the maximum bit rate must be limited so as not to exceed the maximum bit rate according to the target bit rate other than the standard.

In Patent Document 1, rate control is performed with CBR (fixed bit rate) in order to limit the maximum bit rate, and the obtained Q scale is compared with the Q scale of VBR (variable bit rate), and the comparison result is selected. Doing what to do. This CBR rate control is based on TM2 (Test Model) 5 of the MPEG2 standard, but is set at the beginning of the GOP in the calculation of the remaining bit amount of the GOP from which the target bit amount for each frame is obtained The target bit amount generated during 1 GOP is determined based on the maximum bit rate, but the generated bit amount subtracted for each frame is the bit amount actually generated controlled based on the target bit rate. It is.
Usually, since the target bit rate is lower than the maximum bit rate, the amount of remaining bits of the GOP is smaller than the value initially set. Details are shown in FIGS. 10A and 10B described later. As the curves CV11 and CV12 indicated by the solid lines become pictures after the GOP, they gradually become excessive. That is, the target bit amount of one frame obtained at the beginning of the GOP is appropriate, but the target bit amount of one frame gradually increases, and it becomes difficult to limit the maximum bit rate in the second half of the GOP. Go.
Similarly, when the restriction at the minimum bit rate is applied, the same trouble as described above occurs when the CBR rate control based on the TM2 of the MPEG2 standard is performed as shown by the curve CV13 shown by the solid line in FIG.

  The Q scale rate-controlled by the CBR that limits the maximum bit rate takes a smaller value than the VBR Q scale in a normal coding state, and is not selected as a Q scale used for coding. However, as a result of performing the encoding process using the VBR Q scale, when the bit rate exceeds the maximum bit rate, the CBR Q scale increases, the CBR Q scale is selected, and the maximum bit rate Restrictions are made. At this time, in the normal state, the Q scale of the CBR is decreasing, and since the minimum Q scale value is set to 0, the Q scale is stuck to 0, and the maximum bit rate is limited. Since Q scale starts increasing from 0 when it is necessary, it takes time until the VBR Q scale is exceeded, and the bit rate control is delayed.

  In the case of the minimum bit rate, the magnitude relation of the Q scale is reversed. Usually, the Q scale of the CBR becomes a large value, and if the limit is not applied, the Q scale becomes a much larger value to limit the minimum bit rate. Furthermore, the time until the VBR Q scale or less is further increased. During this time, image quality degradation due to the bit rate becoming too low becomes a problem with the minimum bit rate limitation.

  SUMMARY OF THE INVENTION An object of the present invention is to provide an image encoding apparatus and method for overcoming the above-described problems in Patent Document 1 and performing encoding with high image quality without depending on image complexity and changes.

According to the present invention, an image signal to be encoded is encoded by a variable bit rate method and a fixed bit rate method at a response speed, a plurality of target bit rates, a maximum bit rate, and a minimum bit rate given from an external device. An image encoding device comprising:
To achieve the above given above SL plurality of target bit rate when encoding an image signal of the coding target at a variable bit rate system, according to the coding difficulty of the image signal in the first image processing unit First quantization scale calculation means for calculating a plurality of first quantization scales that change
When the image signal to be encoded is encoded by the fixed bit rate method, the second quantum processing unit is smaller than the first image processing unit so as not to exceed the given maximum bit rate. A second quantization scale calculating means for calculating a quantization scale;
When the image signal to be encoded is encoded by a fixed bit rate method, it is smaller than the first image processing unit so as not to fall below the minimum bit rate given according to the image signal to be encoded. Third quantization scale calculating means for calculating a third quantization scale in a third image processing unit;
A larger quantization scale is selected by comparing the first quantization scale and the second quantization scale, and a smaller quantization scale is compared by comparing the selected quantization scale with the third quantization scale. A quantization scale selection means for selecting
Coding means for performing a coding process on the image signal to be coded using the quantization scale selected by the quantization scale selection means;
To the first quantization scale calculating means for calculating said first quantization scale, the given plurality of was in accordance with the coding difficulty when encoding an image signal of the coding target at a variable bit rate system target bit rate and the amount of Coca threshold is set,
The first quantization scale calculation means calculates a plurality of first quantization scales in the first image processing unit in order to achieve each of the set target bit rates.
The quantization scale selection means is at least:
The smaller one by comparing the first quantization thresholds corresponding to the first quantization scale of the first quantization scale and the quantization thresholds of the plurality of first quantization scale first select a value of 1, select the greater by comparing the second quantization scale of the first value and the plurality of first quantization scale and the selected as the second value,
The greater by comparing the second amount coca scale calculated by the second value and the second quantization scale calculation unit described above selects selected as the third value, the third value obtained by the selected selecting smaller as the fourth value is compared with the third quantization scale calculated in the third quantization scale calculation means,
Doing
Outputting the finally selected fourth value to the encoding means;
An image encoding device is provided.

Further, according to the present invention, an image signal to be encoded is transmitted from an external device with a response speed, a plurality of target bit rates, a maximum bit rate, a minimum bit rate, a variable bit rate method and a fixed bit rate method. An image encoding method for encoding, comprising:
To achieve the above given above SL plurality of target bit rate when encoding an image signal of the coding target at a variable bit rate system, according to the coding difficulty of the image signal in the first image processing unit A first quantization scale calculating step of calculating a plurality of first quantization scales that change
When the image signal to be encoded is encoded by the fixed bit rate method, the second quantum processing unit is smaller than the first image processing unit so as not to exceed the given maximum bit rate. A second quantization scale calculating step for calculating a quantization scale;
When the image signal to be encoded is encoded by a fixed bit rate method, it is smaller than the first image processing unit so as not to fall below the minimum bit rate given according to the image signal to be encoded. A third quantization scale calculating step of calculating a third quantization scale in a third image processing unit;
A larger quantization scale is selected by comparing the first quantization scale and the second quantization scale, and a smaller quantization scale is compared by comparing the selected quantization scale with the third quantization scale. A quantization scale selection step of selecting
An encoding step of performing an encoding process on the image signal to be encoded using the quantization scale selected by the quantization scale selection step;
Oite to the first quantization scale calculation step of calculating the first quantization scale, given above in accordance with the coding difficulty when encoding an image signal of the coding target at a variable bit rate system bit rate and the amount Coca threshold of the plurality of targets are set,
In order to achieve each of the plurality of target bit rates set in the first quantization scale calculation step, a plurality of first quantization scales are calculated in the first image processing unit,
In the quantization scale selection step, at least,
More smaller than the first quantization thresholds corresponding to the first quantization scale of the first quantization scale and the upper SL amount Coca threshold among the plurality of first quantization scale was selected as the first value, it selects the greater by comparing the second quantization scale of the first value and the plurality of first quantization scale and the selected as the second value,
The greater by comparing the second amount coca scale calculated by the second value and the second quantization scale calculation step of the above selected selected as the third value, the third value obtained by the selected selecting smaller as the fourth value is compared with the third amount Coca scale calculated above Symbol third quantization scale calculation step,
Doing
The finally selected fourth value is used in the encoding step.
An image encoding method is provided.

  According to the present invention, it is possible to realize an effective encoding suitable for the complexity and ease of an image, and when an easy image is rapidly input after a complex image continues, the easy image continues. After that, even when a complex image is suddenly input, it is possible to perform an effective encoding process without causing deterioration in image quality in the transition region.

In the present invention, by setting an appropriate quantization scale QSC according to an image and a change in the image, a code that is close to the target bit rate on the average according to the encoding difficulty level of the image signal to be encoded It is possible to encode an image signal to be encoded at a bit rate, and to ensure that the maximum bit rate determined by the standard is not exceeded. In other words, by limiting the maximum bit rate and the minimum bit rate of the VBR, image quality degradation at the changing point of the encoding difficulty level and image quality degradation by lowering the bit rate thereafter to cancel that the bit rate is excessively increased Is suppressed.
The reason is that the maximum bit rate limit does not work correctly as the GOP remaining bit amount is obtained by using the actual generated bit amount, for example, and the generation of P and B pictures occurs in the latter half of the GOP. The amount of bits increases, for example, the ratio between I, P, and B pictures of the parameter (global complexity measure) indicating the complexity of the screen used in TM5 step 1 of the MPEG2 standard collapses, and the maximum bit of the I picture in the next GOP Since the target bit amount for limiting the rate is kept small, it is possible to prevent a problem that the picture quality of the I picture is deteriorated.
Furthermore, according to the present invention, the virtual buffer for obtaining the Q scale for limiting the minimum bit rate is not increased indefinitely, and the Q scale is quickly changed when the coding difficulty is changed to an easy image. The image quality can be lowered, and the deterioration of the image quality during this period can be prevented.

Basic Configuration of Image Encoding Device FIG. 4 is a diagram showing an overall basic configuration of an image encoding device according to an embodiment of the present invention.
An image encoding device 1 according to an embodiment of the present invention includes a control circuit 10, an encoder (encoding device) 20, and an image input device 30.
The image input unit of the image input device 30 in the image encoding device 1 is, for example, an imaging device such as a video camera or an image reproduction device that reproduces an image signal from another recording medium. And having a function of converting an analog image signal into a digital signal. In the present embodiment, a digital image signal is input to the image encoding device 1.

  The control circuit 10 outputs control signals S12 and S13 to the image input device 30 and the encoder 20, respectively, according to the control signal S1 input from the outside of the image encoding device 1, and the operation of the image input device 30 and the encoder 20 To control.

For example, the image input device 30 outputs digital image data S30 input from an imaging device such as a video camera to the encoder 20 as image data to be encoded. The image input device 30 generates image information data S32 according to the type and characteristics of the input image signal and supplies the image information data S32 to the control circuit 10.
The control circuit 10 receives the image information data S32 and outputs a compression coding parameter S12 to the encoder 20 according to the characteristics of the input image signal. The compression coding parameter S12 includes a plurality of parameters such as a target coding rate, a maximum coding rate, a variable bit rate instruction signal, or these signals, which will be described later.

The encoder 20 converts the digital image data S30 sent from the image input device 30 in accordance with a compression encoding parameter S12 from the control circuit 10 based on a predetermined encoding algorithm, for example, an encoding algorithm based on the MPEG2 standard. The bit stream S20 is output after compression encoding.
The encoder 20 appropriately selects a variable encoding rate and a fixed encoding rate to obtain an optimal encoding rate in accordance with image characteristics and image changes, and based on the encoding rate (Q scale). In accordance with the complexity of the image and changes in the image, the image quality of the entire image is made uniform and the compression efficiency is improved.
As a result of the image data being compression-encoded by the encoder 20, the data amount of the bit stream S20 is much reduced compared to the image data S30 before the compression encoding. For this reason, by transmitting and accumulating the bit stream S20 compression-encoded by the encoder 20, the transmission capacity of the transmission path or the storage capacity of the image recording medium can be greatly reduced.

Basic Configuration of Encoder (Encoding Device) FIG. 5 is a block diagram showing the basic configuration of the encoder 20 in this embodiment.
The encoder 20 includes a variable coding bit rate (VBR) control circuit 22a, a fixed coding bit rate (CBR) control circuit 23a, a quantization (Q) scale selection circuit 25a, and an image compression coding unit 24.
The fixed bit rate control circuit 23a includes, for example, a maximum CBR control unit 23A and a minimum CBR control unit 23B as illustrated in FIG.
The VBR control circuit 22a is a first quantization scale setting unit that sets a first quantization scale (first Q scale) for performing image coding in a first image processing unit, for example, a GOP unit including a plurality of pictures. Yes, the first quantum that changes slowly according to the degree of difficulty of encoding the input image signal (image data S30) to be encoded from the image input device 30 in the first image processing unit with respect to the target bit rate. Set the quantization scale QS1.
The CBR control circuit 23a uses the first image processing unit, for example, a second image processing unit shorter than the GOP unit, for example, a picture or macro block (MB) unit, and specifies the maximum specified by the sequence header at the beginning of the bitstream. Second and third quantization scale setting means for setting second and third quantization scales QS2 and QS3 that change rapidly so as not to exceed the maximum bit rate with respect to the bit rate. Therefore, the fixed bit rate control circuit 23a includes, for example, a maximum CBR control unit 23A and a minimum CBR control unit 23B as described with reference to FIG.
The quantization scale selection circuit 25a compares the first quantization scale QS1 set by the VBR control circuit 22a with the second and third quantization scales QS2 and QS3 set by the CBR control circuit 23a, and compares these quantum values. An appropriate Q scale is output to the image compression encoding unit 24 as a quantization scale QSC for performing an encoding process on the input image signal (image data S30) to be encoded.
The image compression encoding unit 24 encodes the image data S30 based on the quantization scale QSC from the quantization scale selection circuit 25a and the compression encoding parameter S12 from the control circuit 10.

First Embodiment of Encoder FIG. 6 is a block diagram showing the configuration of the first embodiment of the encoder 20 illustrated in FIG.
The encoder 20 of the first embodiment includes a VBR buffer recording unit 21, a VBR control unit 22, a maximum CBR control unit 23A, a minimum CBR control unit 23B, an image compression encoding unit 24, and a quantization scale comparison unit 25.
The VBR control circuit 22a illustrated in FIG. 5 corresponds to the VBR control unit 22, the CBR control circuit 23a corresponds to the maximum CBR control unit 23A and the minimum CBR unit 23B, and the quantization scale selection circuit 25a corresponds to the quantization scale comparison unit 25. The image compression encoding unit 24 corresponds to the image compression encoding unit 24 as it is. In addition, the encoder 20 of the first embodiment includes a VBR buffer recording unit 21.

Compared with the encoder 20 of the first embodiment and the encoder (encoding device) described in Patent Document 1, the CBR control unit 23 in Patent Document 1 is different from the maximum CBR control unit in the encoder 20 of the first embodiment. 23A and the minimum CBR part 23B.
In addition, details will be described with reference to FIG.

Image Compression Encoding Unit FIG. 7 is a diagram showing a configuration of the image compression encoding unit 24 in the encoder 20 illustrated in FIG.
The image compression encoding unit 24 is input from the image input device 30 according to the target bit rate BRobj, the maximum bit rate BRmax, the VBR instruction signal VBD, and the quantization scale QSC provided from the quantization scale comparison unit 25. The image data S30 to be encoded is compressed and encoded, and a bit stream S20 is output.
Further, the image compression encoding unit 24 generates a generated bit amount BQT indicating the bit amount generated by the compression encoding and a PCT signal indicating the image encoding type, and as illustrated in FIG. To the maximum CBR control unit 23A and the minimum CBR control unit 23B.

  The image compression encoding unit 24 includes an adder (adder / subtractor) 201, a discrete Coisan transform processing unit (DCT (Discrete Cosine Transform) processing unit) 202, a quantization unit 203, a variable length encoding unit (VLC) 204, an inverse quantum. A conversion unit 205, an inverse DCT unit 206, an adder 207, a frame memory 208, a motion detection unit 209, a motion compensation unit 210, a switching circuit 211, and an encoding control unit 212.

The digital image signal S30 to be encoded input from the image input device 30 to the image compression encoding unit 24 is obtained by the adder 201 to obtain a difference from the previous reproduced image signal input from the switching circuit 211. Is output to the DCT unit 202.
In the intra picture (I picture) mode, since no significant image is input from the switching circuit 211 (in FIG. 7, the output from the switching circuit 211 in this mode is set to 0 of the ground potential), it is input to the adder 201. The processed image data is output to the DCT unit 202 as it is.
In the motion compensation prediction mode, since the image data from the motion compensation unit 210 based on the previous reproduced image is input to the adder 201 via the switching circuit 211, the difference is obtained by the adder 201. The obtained difference is input to the DCT unit 202.

The DCT unit 202 performs discrete cosine transform on the image signal input from the adder 201 to reduce the spatial redundancy of the image data. The transform coefficient obtained by DCT is input to the quantization unit 203.
The quantization unit 203 quantizes the DCT transform coefficient obtained by the DCT unit 202 according to the quantization scale QSC from the quantization scale comparison unit 25 input via the encoding control unit 212, and performs quantization DCT. The transform coefficient is output to variable length coding section 204 and inverse quantization section 205. As described above, the quantization scale in the quantization unit 203 is controlled by the encoding control unit 212 that receives the quantization scale QSC from the quantization scale comparison unit 25.

  The variable length coding unit 204 performs variable length coding on the DCT transform coefficient quantized by the quantization unit 203, generates a coded bit string, that is, a bit stream S20, and sequentially outputs it.

The encoding control unit 212 controls the quantization rate (Q scale) of the quantization unit 203, recording of image data in the frame memory 208, switching of the switching circuit 211, and the like. As shown in FIG. 7, for example, the encoding control unit 212 receives the target bit rate BRobj, the maximum bit rate BRmax, the VBR instruction signal VBD input from the control circuit 10, and the quantum from the quantization scale comparison unit 25. A predetermined quantization scale is set in the quantization unit 203 according to the quantization scale QSC, and the quantization rate in the quantization unit 203 is controlled.
The DCT transform coefficient quantized by the quantization unit 203 is input to the inverse quantization unit 205. The quantized DCT transform coefficient is inversely quantized by the inverse quantization unit 205 and restored to the DCT transform coefficient before being input to the quantization unit 203, and further, the inverse DCT unit 206 performs inverse discrete cosine transform (inverse). DCT) is performed and inversely converted into an output image of the adder 201 before being input to the DCT unit 202.
In the mode for performing motion compensation prediction, the adder 207 obtains the image signal subjected to inverse DCT in the inverse DCT unit 206 and the motion compensation prediction in the motion compensation unit 210 output via the switching circuit 211. The added image signals are added back to an image signal equivalent to the original image signal input to the adder 201 and recorded in the frame memory 208.
In the intra picture mode, the adder 207 does not receive an image based on motion compensation prediction in the motion compensation unit 210 via the switching circuit 211, and the image signal subjected to inverse DCT in the inverse DCT unit 206 is recorded in the frame memory 208 as it is. Is done.
When motion compensation prediction is performed in the motion detection unit 209 and the motion compensation unit 210, an image signal recorded in the frame memory 208 is used. That is, based on the image signal recorded in the frame memory 208, the motion detection unit 209 performs motion detection from the next image source signal to detect a motion vector. Further, based on the motion vector, the motion compensation unit 210 performs motion compensation prediction. The image signal obtained by the motion compensation prediction is output to the adder 201 via the switching circuit 211, and the adder 201 obtains a difference from the next input image source signal. Similarly, in the mode for performing motion compensation prediction, an image signal obtained by motion compensation prediction output from the switching circuit 211 is input to the adder 207 and corresponds to the image input to the adder 201 in the adder 207. Find the image you want.

In this manner, the input digital image data S30 is encoded by the image compression encoding unit 24, and the bit stream S20 is generated.
When the VBR instruction signal VBD supplied from the control circuit 10 is 0, the target bit rate BRobj is output as the bit rate of the sequence header in the bit stream S20. When the VBR instruction signal VBD is 1, the maximum bit rate BRmax is used as the bit rate. Is output.
When a quantization scale is set for each macroblock (MB), the image compression coding unit 24 uses this as a basis for the quantization scale aimed at visual effects based on step 3 of coding rate control in TM5 of MPEG2. This is used for quantization. Then, a bit stream S20 generated for each MB is output.

  The image compression encoding unit 24 of the present embodiment forcibly cuts the DCT coefficient when the encoding rate control by the quantization scale is not successful and the VBV buffer unit (not shown) is likely to underflow. For example, it has a general function as an ordinary MPEG2 image encoding device.

  The above-described image compression encoding unit 24 realizes encoding of an image signal at a variable encoding rate by one-time encoding processing, that is, one-pass processing, instead of the above-described two-stage encoding processing method. it can. For this reason, for example, in an image processing apparatus that captures and records images in real time, such as a video camera, it is possible to realize image encoding at a variable encoding rate with a single encoding process, and image signals having different characteristics. The image quality can be made uniform, and the encoding efficiency can be improved by the variable encoding rate control.

VBR buffer recording unit The VBR buffer recording unit 21 holds the VBR buffer value VBRb, and the total value in the series of encoding processes in the image compression encoding unit 24 described above reduces the deviation from the target bit rate BR _OBJ. Is provided to do.
Before starting the encoding process, the VBR buffer recording unit 21 stores the VBR buffer initial value BR _INI input from the control circuit 10. As the simplest initialization method, the VBR buffer recording unit 21 is initialized according to the initialization control signal from the control circuit 10 and the VBR buffer initial value BR _INI is set to “0” before starting the processing. , Set to a predetermined value.
The VBR buffer recording unit 21 provides the VBR buffer initial value BR _INI recorded in the VBR control unit 22 after compression encoding is started in the encoder (encoding device) 20, and VBR control is performed at the end of encoding in the encoder 20. The VBR buffer value VBRb is input from the unit 22 and recorded. The VBR buffer recording unit 21 outputs the recorded VBR buffer value VBRb to the VBR control unit 22 in the next encoding process.

VBR Control Unit The VBR control unit 22 obtains the VBR quantization (Q) scale VBRq according to the VBR response speed VBRr, the target bit rate BRobj, the maximum bit rate BRmax, and the minimum bit rate BRmin input from the control circuit 10. To the quantization scale comparison unit 25.
VBR control section 22, the target amount of bits generated 1GOP obtained from the target bit rate BR _OBJ to compare actual bit amount generated in the GOP, if the bit amount than the target bit rate BR _OBJ too occurs For example, the Q scale is gradually increased to gradually suppress the generated bit amount, and conversely, if the bit amount is not generated more than the target bit rate BR _{OBJ, the} Q scale is gradually decreased to gradually increase the generated bit amount. . As a result, the Q scale does not fluctuate with a good response, so the VBR rate control becomes a generated bit amount corresponding to the encoding difficulty level, and the Q scale is gradually changed according to the generated bit amount. The generated bit amount corresponds to the target bit rate BR _OBJ .
In the VBR control unit 22, as internal parameters, the VBR buffer value VBRb for obtaining the VBR quantization scale VBRq, the remaining bit amount RBG in the GOP when encoding with the target bit rate BR _OBJ , and the response speed of the VBR quantization scale VBRq VBR response speed VBRr representing is held.

In the VBR control unit 22, the VBR quantization scale VBRq is obtained by the following equation.

VBRq = VBRb / (2 ⁿ × VBRr) (1)
In equation (1), n is a calculation parameter, which is set according to the application, for example, n = 9.

When the VBR quantization scale VBRq becomes 1, the generated bit amount increases rapidly and the VBR rate control becomes unstable. Therefore, when the VBR quantization scale VBRq calculated by the equation (1) is 1 or less, the VBR control The unit 22 takes measures such as forcibly changing the VBR quantization scale VBRq to 2.
Further, paying attention to the difference between the I / P / B pictures in the GOP, the VBR control unit 22 sets the coefficient, for example, the value obtained by the equation (1) when the coding type is the B picture, for example, A value obtained by multiplying 1.4 is defined as a VBR quantization scale VBRq. That is, in the case of B pictures, the VBR quantization scale is always set to 1.4 times the VBR quantization scale of I and P pictures. The reason for this is that when the B picture is encoded somewhat coarsely compared to the I and P pictures, and the coding amount that can be saved in the B picture is allocated to the I and P pictures. This is because the S / N is improved, and the S / N of the B picture generated with reference to this is also improved, so that the S / N can be improved comprehensively.
In the VBR control unit 22, the Q scale of the I / P / B picture is determined by a simple method as described above. However, the VBR control unit 22 is not limited to the above-described example. You can also add a difference in Q scale.

In the VBR control unit 22, a target bit amount is set for each GOP as the remaining bit amount RBG in the GOP.
When the VBR control unit 22 receives the generated bit amount for each MB (macroblock), for example, the VBR control unit 22 subtracts that amount (target bit amount for each GOP). Since the deviation from the target bit amount remains in the remaining bit amount RBG in the GOP at the end of each GOP, the VBR control unit 22 subtracts this (remaining bit amount RBG) from the VBR buffer value VBRb. At this time, if the VBR buffer value VBRb becomes negative, the VBR control unit 22 sets it to 0. When the VBR buffer value VBRb is likely to overflow, the VBR control unit 22 fixes it (VBR buffer value VBRb) to the maximum value.

As described above, the initial value of the VBR buffer value VBRb at the start of the encoding process is input from the VBR buffer recording unit 21 to the VBR control unit 22. Here, for example, when the VBR buffer value VBRb input from the VBR buffer recording unit 21 is other than 0, the VBR control unit 22 uses the value, and when 0, the value obtained by the following equation is used.

VBRb = m · VBRr / (BRobj × 2 ^k ) (2)
In Equation (2), m and k are calculation parameters. Here, for example, m = 2.5 × 10 ⁷ and k = 10.

  A value supplied from the control circuit 10 is used as the response speed VBRr in the equation (2). For example, when the response speed VBRr is set to a value of about 200, several tens of seconds can be obtained when the response speed VBRr is set to a value of about tens of seconds and 4000. The VBR buffer value VBRb calculated by Equation (2) at the end of encoding is transmitted to the VBR buffer recording unit 21 and stored.

For the remaining bit amount RBG in the GOP, a value obtained by the following equation is first set.

RBG = BRobj · N / PCR (3)
In Equation (3), N is the I picture interval, and PCR is the picture rate.

CBR Control Unit The CBR control unit according to the first embodiment includes a maximum CBR control unit 23A and a minimum CBR control unit 23B.
FIG. 8 is a diagram illustrating a relationship between the configuration of the maximum CBR control unit 23A and the minimum CBR control unit 23B and the image compression encoding unit 24, particularly, the encoding control unit 212 according to the first embodiment.
The maximum CBR control unit 23A includes a GOP allocated bit amount calculation unit 2311, a GOP remaining bit amount buffer 2312, a TM5 step 1 calculation unit 2313, and a TM5 step 2 calculation unit 2314.
The minimum CBR control unit 23B includes a GOP allocated bit amount calculation unit 2321, a GOP remaining bit amount buffer 2322, a TM5 step 1 calculation unit 2323, and a TM5 step 2 calculation unit 2324.
In the maximum CBR control unit 23A and the minimum CBR control unit 23B, in this embodiment, for example, CBR rate control is performed based on a TM5 algorithm for MPEG2. Of course, in the image encoding device 1 according to the embodiment of the present invention, the rate control in CBR is not limited to being performed based on the TM5 algorithm, and may be performed based on other control methods. Good. However, the case of performing based on the TM5 algorithm will be described below.

FIG. 9 shows a configuration example of the CBR control unit 23 of Patent Document 1. Similar to the maximum CBR control unit 23A, the CBR control unit 23 includes a GOP allocated bit amount calculation unit 2311, a GOP remaining bit amount buffer 2312, a TM5 step 1 calculation unit 2313, and a TM5 step 2 calculation unit 2314.
A minimum CBR control unit 23B not described in Patent Document 1 is added to the CBR control unit of the embodiment of the present invention.
The maximum CBR control unit 23A and the CBR control unit 23 have similar configurations with respect to the individual components, but the content described in the present embodiment and Patent Document 1 is related to the image compression encoding unit 24, and The processing contents in the CBR control unit 23 are different as described in detail below.

The CBR control unit 23 described in Patent Document 1 illustrated in FIG. 9 performs an operation for calculating the remaining GOP bit amount when obtaining CBRmaxq or when obtaining a Q scale using normal CBR. The generated bit amount for each frame output from the conversion unit 24 is subtracted from the remaining GOP bit amount.
According to the CBR control unit 23 that limits the maximum bit rate, as shown by thick solid line curves CV11, CV12, and CV13 in FIGS. The value obtained from the rate is set. Although the target bit amount of each frame obtained from this is a larger value than that of normal CBR, the encoding process corresponding to the value is not performed, but the rate control is performed by VBR corresponding to the target bit rate BR _OBJ. Therefore, the remaining GOP bit amount changes as shown by the thick curve CV12 in FIG. 10B, and takes a large value even at the end of the GOP. As a result, the target bit amount of one frame becomes a larger value after the GOP, and the Q scale obtained from this becomes not appropriate.
Since it is necessary to adjust the subsequent generated bit amount according to the bit amount generated so far, it is necessary to subtract the actually generated bit amount from the remaining GOP bit amount, but the maximum bit rate or the minimum bit rate of the VBR Since the Q scale of CBR obtained for limiting the value is obtained as a target value regardless of the amount of generated bits in the past, there is no problem even if such a method is used.
This problem also applies to CBR that limits the minimum bit rate.

Ideally, the remaining GOP bit amount should change uniformly at a predetermined rate as shown by thin dotted curve lines CV22 and CV23 in FIGS.
The COP control method of the GOP of the present embodiment using the maximum CBR control unit 23A and the minimum CBR control unit 23B is shown in FIG. 10 (B) and FIG. 10 (C), as indicated by dotted curve CV22 and CV23. The target bit amount R of 1 GOP corresponding to the GOP is set as the remaining GOP bit amount, the encoding is controlled by the target bit amount R of each frame obtained from this, and the bit amount corresponding to the target bit rate BR _OBJ is generated every frame Then, by subtracting the target bit amount of each frame from the remaining GOP bit amount, the remaining GOP bit amount becomes close to zero at the end of the GOP.
According to the CBR control unit of the present embodiment having the maximum CBR control unit 23A and the minimum CBR control unit 23B, the remaining GOP bit amount is almost as shown by the dotted curves CV22 and CV23 in FIGS. In order to change at a constant rate, the bit amount subtracted from the remaining GOP bit amount for each frame is not the generated bit amount controlled based on the target bit rate BR _OBJ but the maximum bit rate BRmax or the minimum bit rate BRmin. The target bit amount for each frame is subtracted.

The maximum CBR control unit 23A according to the present embodiment performs a calculation process of the remaining GOP bit amount when obtaining CBRmaxq. That is, in the present embodiment, each frame output from the encoding control unit 212 as in the relationship between the CBR control unit 23 of FIG. 9 and the encoding control unit 212 of the image compression encoding unit 24 described in Patent Document 1. The target bit amount R output from the TM5 step 1 calculation unit 2313 is changed so as to be subtracted from the remaining GOP bit amount instead of the generated bit amount. Thus, CBRmaxq output regardless of the frame position in the GOP is always an appropriate value.
The same applies to the determination of CBRminq in the minimum CBR control unit 23B in the present embodiment.

The maximum CBR control unit 23A of the present embodiment sets the maximum CBR quantization scale CBRmaxq according to the maximum bit rate BRmax input from the control circuit 10, and outputs the set CBRmaxq to the quantization scale comparison unit 25.
The GOP allocated bit amount calculation unit 2311 of this embodiment determines the bit amount allocated to the GOP from the number of pictures and the bit rate included in each GOP. The remaining amount ΔR (positive or negative) of the bit amount in the GOP before the currently processed GOP is not added to the current processing.
In this embodiment, the target bit amount for a picture to be encoded is determined from the remaining GOP bit amount R, the encoding difficulty level for each picture type (I / B / P), and the number of remaining pictures for each picture type. Is done. Furthermore, the target bit amount for the picture to be encoded is subtracted from the remaining GOP bit amount to calculate a bit amount residual ΔR.

In the TM5 step 1 operation unit 2313 and the TM5 step 2 operation unit 2314 of the present embodiment that performs the processing of steps 2 and 2 of TM5 of the MPEG2 standard, image encoding processing is performed in GOP units.
In processing in units of macroblocks (MB), when a picture is encoded in units of MB and the generated bit amount is likely to exceed the target bit amount in the picture, the Q scale is set large, and conversely the generated bits If the amount is likely to be below the target bit amount, the Q scale is set small. Furthermore, in consideration of the fact that human visual characteristics are sensitive to image quality degradation in flat areas in the screen, the Q scale is set to be small for relatively flat image areas, and conversely, the Q scale is set for fine image areas. It is set large.

In TM2 of MPEG2, the target bit amount R of the next GOP is determined for each GOP. At the end of a GOP, the amount of bits generated in that GOP is a value subtracted from R. This is expressed as ΔR as the remaining bit amount of the GOP. This remaining bit amount ΔR is added to the target bit amount R of the next GOP.
In the MPEG2 TM5, the target bit amount R of the next GOP is determined for each GOP. The maximum CBR control unit 23A and the minimum CBR control unit 23B perform the following different processes to perform the target bit amount R of the next GOP. To decide.

(A) In the TM5 step 1 calculation unit 2313, when the VBR instruction signal VBD = 0 supplied from the control circuit 10, the target bit amount Ri + 1 of the next GOP is determined according to the current GOP bit remaining amount ΔRi according to the following equation. Is calculated. Here, i indicates the GOP number.

Ri + 1 = BRobj.N / PCR + .DELTA.Ri (4)
In equation (4), N is the interval between I pictures, and PCR is the picture rate.

(B) In the TM5 step 1 calculation unit 2313, when the VBR instruction signal VBD = 1 and the bit remaining amount ΔRi of the immediately preceding GOP is negative (ΔRi <0), the target bit amount Ri of the next GOP is calculated by the following equation +1 is calculated.

Ri + 1 = BRmax.N / PCR + .DELTA.Ri (5)

(C) In the TM5 step 1 operation unit 2313, when the VBR instruction signal VBD = 1 and the bit remaining amount ΔRi of the immediately preceding GOP is positive or 0 (ΔRi ≧ 0), the target bit of the next GOP is obtained by the following equation The quantity Ri + 1 is calculated.

Ri + 1 = BRmax · N / PCR (6)

In the case of (a), the processing is the same as the normal MPEG2 TM5 processing by the CBR control unit 23 of FIG. In order not to deviate significantly from the target bit rate BRobj, the bit remaining amount ΔR for each GOP is always fed back to the target bit amount of the next GOP, and the target bit amount Ri + 1 of the next GOP becomes the bit remaining amount ΔRi of the immediately preceding GOP. Will be modified accordingly.
In the case of (b, c), VBR control is performed in the encoder 20, and the target bit of the next GOP according to the remaining bit amount ΔR for each GOP so as not to exceed the maximum bit rate BRmax / minimum bit rate BRmin. A quantity is calculated. When the bit remaining amount ΔRi of the immediately preceding GOP is positive or 0, the bit remaining amount ΔRi of the immediately preceding GOP is not fed back for the reason described above, and the target bit amount Ri + 1 of the next GOP is expressed by equation (6). The maximum bit rate BRmax / minimum bit rate BRmin is determined.

  The maximum CBR control unit 23A and the minimum CBR control unit 23B obtain a maximum CBR quantization scale CBRmaxq and a minimum CBR quantization scale CBRminq based on the target bit amount of GOP calculated as described above, and a quantization scale comparison unit 25.

Quantization scale comparison unit The quantization scale comparison unit 25 receives the VBR quantization scale VBRq input from the VBR control unit 22, the maximum CBR quantization scale CBRmaxq input from the maximum CBR control unit 23A, and the minimum CBR control unit 23B. The base quantization scale QSC is calculated according to the minimum CBR quantization scale CBRminq, and provided to the image compression encoding unit 24.

  In the quantization scale comparison unit 25, as shown in the flowchart of FIG. 11, in step 1 (S1), VBRq and CBRminq are compared and the smaller one is selected. As a result, the maximum value of the Q scale is limited so that the Q scale is not too large and the bit rate does not decrease too much. Next, in step 2 (S2), the quantization scale comparison unit 25 compares the comparison result with CBRmaxq and selects the larger one. This limits the minimum value of the Q scale so that the Q scale is not too small and the bit rate is not increased too much.

  For example, as shown in FIG. 12A, the CBR Q scale (CBRminq) for limiting the minimum bit rate BRmin normally takes a large value, but when the limitation by the minimum bit rate BRmin is necessary, the QBR Qscale A value less than the scale VBRq and a CBRminq that is a Q scale smaller than VBRq are selected and used for encoding in the image compression encoding unit 24, so that the bit rate does not become too small due to a large Q scale. Bit rate is limited.

When a normal image or a difficult image continues to be input, CBRminq takes a large value. However, the maximum value of CBRminq is converted into a code in MPEG, for example, as illustrated in FIGS. Since the upper limit of “31” is set, “31” is never exceeded. However, when CBRminq is obtained by control based on TM5 of the MPEG2 standard, it is obtained by multiplying the value of the corresponding virtual buffer by the value of (31 / reaction parameter), and if this is converted to a code and exceeds “31”, “ 31 "is calculated. Since the value of the virtual buffer is not limited in TM5, when a normal state or a difficult image is continuously input, CBRminq continues to increase as indicated by a broken line in FIG.
Even if the encoding difficulty of the input image data to be encoded becomes easy in this state, the value of the virtual buffer becomes sufficiently small, and it takes time until the Q scale CBRminq to be calculated below VBRq. During this period, the image quality deteriorates.
In order to deal with such a problem, as illustrated in FIG. 12B, if CBRminq exceeds VBRq for a short period, for example, every frame, CBRminq is reset to a value close to VBRq.

As a method of setting CBRminq to a value close to VBRq, for example, the value of the virtual buffer is rewritten to (VBRq × reaction parameter) / 31. Accordingly, the period during which the image quality degradation in the transition regions Z1 and Z2 shown in FIG. 3 is shortened to, for example, several slices, and the Q scale between them is not too large. Image quality degradation when switching to an easy image is suppressed.
As a result, the change of the Q scale in the transition regions Z1 and Z2 shown in FIG. 3 described above becomes like the curve shown in FIG. 13, and the problem that occurs at the change point of the encoding difficulty (transition regions Z1 and Z2) is Solved.
As described above, according to the first embodiment, by setting an appropriate quantization scale QSC according to an image and a change in the image, the target bit rate is averaged according to the encoding difficulty level of the input image signal. It is possible to encode the input image signal at a code bit rate close to, and to ensure that the maximum bit rate determined by the standard is not exceeded.

  The same applies to CBRmaxq, but the virtual buffer used as the GOP remaining bit amount buffer 2322 often uses an unsigned type argument, and when it becomes a negative value, it is suppressed to zero, so the effect is small compared to CBRminq. There are many cases, and the Q scale does not immediately become a large value, that is, it seems that there is no problem because the image quality encoded for a short period of time becomes too good, but CBRmaxq is similar to CBRminq. The value may be reset in a short period of time.

Overall Processing of Image Encoding Device FIG. 14 is a flowchart showing the overall operation of the image encoding device according to the first embodiment.
Hereinafter, the operation of the image encoding device 1 according to the first embodiment will be described using specific numerical examples of the respective parameters.

In step 11, after the control circuit 10 starts the encoding process, first, the VBR control unit 22 obtains an initial value of the VBR buffer value VBRb according to the equation (2) according to the given parameter.
Here, for example, BRobj = 4 Mbps, maximum bit rate BRmax = 8 Mbps, VBR response speed VBRr = 300, m = 2.5 × 10 ⁷ in equation (2), and k = 10, 2), the VBR buffer value VBRb becomes (2.5 × 10 ⁷ × 300 / (4 × 10 ⁶ × 2 ¹⁰ )) = 1920000.

In step 12, the VBR control unit 22 obtains the target bit amount of GOP based on the equation (3).
In Equation (3), N = 15 when GOP is in units of 15 pictures. For example, when an NTSC television image signal is encoded, each frame is updated at 29.97 Hz. In this case, the picture rate PCR is 29.97. For this reason, the target bit amount of the GOP is (4 × 10 ⁶ × 15 / 29.97) = 2002002 according to Equation (3).

In step 13, the control circuit 10 obtains the target bit amount of CBR.
For example, the target bit amount of CBR is (8 × 10 ⁶ × 15 / 29.97) = 4004004 according to the maximum bit rate BRmax, the number N of GOP pictures, and the picture rate PCR.

In step 14, based on the VBR buffer value VBRb obtained in step 11, the VBR control unit 22 obtains the VBR quantization scale VBRq using equation (1) and outputs it to the quantization scale comparison unit 25.
When n = 9 in the formula (1), the VBR quantization scale VBRq is (1920000 / (2 ⁹ × 300)) = 12.5. Since the quantization scale is required to be an integer value, for example, VBR quantization scale VBRq = 13 is set here.
When the picture coding type is 1 or 2, ie, an I or P picture, the variable quantization scale VBRq = 13 is set. When the picture coding type is 3, ie, a B picture, the variable quantization scale is The quantization scale calculated in (1) is set to a value obtained by multiplying by 1.4. That is, in the case of a B picture, VBRq = 18 is set.

  In step 15, the maximum CBR control unit 23A obtains the maximum CBR quantization scale CBRmaxq according to the maximum bit rate BRmax, and the minimum CBR control unit 23B obtains the minimum CBR quantization scale CBRminq according to the minimum bit rate BRmin. To the scaled scale comparison unit 25.

  In step 16, the quantization scale comparison unit 25 performs the quantization scale determination process described with reference to FIG. That is, the VBR quantization scale VBRq and the minimum CBR quantization scale CBRminq are compared, the smaller one is selected, the selection result is compared with the maximum CBR quantization scale CBRmaxq, and the larger one is selected, and the quantization scale is selected. The QSC is output to the encoding control unit 212 of the image compression encoding unit 24.

  In step 17, the image compression coding unit 24 performs coding processing on the input image signal in units of 1 MB (1 macroblock) with the quantization scale provided by the quantization scale comparison unit 25.

  In step 18, the maximum CBR control unit 23A and the minimum CBR control unit 23B subtract the target bit amount from the set GOP target bit amount.

In step 19, the control circuit 10 performs control so that the processes from step 15 to S18 are repeated until the encoding process for all image signals of 1 GOP is completed.
After the 1 GOP encoding process is completed, the remaining bit amount RBG in 1 GOP is extracted in step 20.
In step 21, the VBR control unit 22 updates the VBR buffer value in accordance with the remaining bit amount RBG.

In step 22, the maximum CBR control unit 23 </ b> A and the minimum CBR control unit 23 </ b> B calculate the GOP bit remaining amount ΔR according to the generated bit amount, and according to the bit remaining amount, the maximum CBR based on Expression (6). The target bit amount of GOP in the control unit 23A and minimum CBR control unit 23B is updated.
As described above, when the bit remaining amount ΔR of the previous GOP is positive or 0, the target bit amount of the next GOP is calculated based on the equation (6).

  In step 23, the control circuit 10 determines whether or not the image encoding process is completed. If the encoding process continues, the process returns to step 14 and the encoding process in the next GOP is performed.

  As described above, in the image encoding device 1, since the processes from step S14 to S23 are repeatedly performed, the encoding process is performed for each GOP for the digital image data S30 to be encoded.

The above-described processing is not limited to GOP units, and encoding processing can be performed in other image processing units, for example, picture units.
In addition, each step in the encoding process can be realized in a different order, and other processing steps can be added as necessary. For example, in several pictures or several GOPs immediately after the start of processing, the input image signal is first encoded at a fixed encoding rate set according to the target bit rate BRobj by the maximum CBR control unit 23A and the minimum CBR control unit 23B, It is also possible to set parameters for encoding processing in accordance with the processing results in these pictures or GOPs, and then shift to encoding processing at a variable encoding rate.

An image encoding process when the encoding difficulty level of the input image signal to be encoded is changed will be described.
First, when the input image for a while after the start of the encoding process is encoded using the quantization rates 13 and 18 calculated in step 12, the encoding is performed so that the bit amount is equal to or less than the maximum bit rate BRmax. It shall have difficulty.
In the encoding process, for example, the maximum CBR control unit 23A aggregates the generated bit amount for each MB (macroblock), and subtracts the generated bit amount from the target bit amount R for each picture to obtain the remaining bit amount. ΔR is calculated. If it is used in common by several blocks, it may be aggregated in another block and the aggregated result may be obtained.
Then, the maximum CBR control unit 23A and the minimum CBR control unit 23B obtain the CBR quantization scale so that the remaining bit amount ΔR obtained by subtracting the target bit amount approaches zero.
As a result of the CBR quantization scale being lowered, the quantization scale QSC determined by the quantization scale comparison unit 25 according to the determination process of FIG. 11 is provided to the encoding control unit 212 of the image compression encoding unit 24. At the end of the GOP quantization process, for example, a large remaining bit amount ΔR may be left.

In this embodiment, since the bit remaining amount ΔR is positive, feedback processing based on the remaining amount of bits is not performed, and the target bit amount R of the next GOP is based on the maximum bit rate BRmax as shown in Expression (6). Desired.
On the other hand, in the VBR control unit 22, the bit amount generated for each MB is subtracted from the bit remaining amount RBG of the GOP calculated based on the equation (3). In response to this, the VBR quantization scale VBRq in the next GOP is obtained by Expression (2). Thus, the VBR quantization scale changes for each GOP, and the generated bit amount can be prevented from deviating greatly from the target bit amount.

  It is assumed that image data with a high degree of difficulty of encoding that exceeds the maximum bit rate BRmax is input with the VBR quantization scale obtained above. In this case, the maximum quantization scale CBRmax q of the maximum CBR control unit 23A becomes larger during one picture and exceeds the VBR quantization scale VBRq. Therefore, the quantization scale comparison unit 25 performs the process illustrated in FIG. The larger VBRq is selected and provided to the image compression encoding unit 24 as the quantization scale QSC.

As described above, when an input image is encoded at a variable bit rate by a single encoding process, it must be ensured that the maximum bit rate is not exceeded when the difficulty of encoding the image suddenly increases.
It is desirable that the bit rate does not fall below the minimum bit rate even when there is a sudden change from a difficult image to an easy image.
In the present embodiment, the CBR control unit normally provided in the image encoding device is diverted to the maximum CBR control unit 23A and the minimum CBR control unit 23B, so that the configuration of the image encoding device is not significantly changed. Rate control can be realized.
Further, when the maximum CBR control unit 23A and the minimum CBR control unit 23B are not used as in the present embodiment, the VBV buffer is likely to underflow, so that the DCT coefficient is cut by the image compression encoding unit 24. Therefore, it is expected that the image quality can be improved according to the present embodiment.

  At the end of encoding, the VBR buffer value VBRb is output to the VBR buffer recording unit 21 and held by the VBR buffer recording unit 21. Since this held value becomes the initial value of the VBR buffer value VBRb at the start of the next encoding, the VBR buffer value VBRb becomes a value at the end of the previous encoding process.

FIG. 13 is a diagram illustrating a quantization scale of image encoding processing in the image encoding device 1 of the present embodiment. When a complex image is input, when the encoding bit rate is likely to exceed the maximum bit rate determined by a predetermined image encoding standard, the CBR control in the maximum CBR control unit 23A starts to function, and the encoding rate is It is controlled not to exceed the maximum bit rate.
In the case of an easy image, it depends on the CBR control function of the minimum CBR control unit 23B.

  With such control, the image coding apparatus 1 according to the embodiment of the present invention keeps the maximum coding rate determined by the standard, and according to the degree of difficulty in coding the image, and rapidly increases the complexity of the image. Even if there is a slight change, the image quality can be kept almost uniform by appropriately controlling the bit rate.

The characteristics of the image encoding device 1 according to the first embodiment of the present invention are summarized as follows.
(1) In the image encoding apparatus 1 according to the first embodiment of the present invention, in particular, in the VBR rate control performed by operating the Q scale, the maximum value of the bit rate between the maximum CBR control unit 23A and the minimum CBR control unit 23B. And the minimum value is added. The maximum value and the minimum value of the bit rate are limited by performing rate control by a constant bit rate (CBR) method, obtaining the respective Q scales, and comparing the size of the VBR with the Q scale by the quantization scale comparison unit 25. An appropriate one is selected, and the selected quantization scale QSC is used for encoding in the image compression encoding unit 24.
(2) When rate control is performed in the CBR system, for example, update calculation is performed on the bit amount R allocated to an uncoded picture in a GOP generally used in step 1 of TM5 of the MPEG2 standard. The bit amount for each picture subtracted by is not the actually generated bit amount used in the normal CBR system, but the target bit amount calculated before encoding.
(3) When rate control is performed using the CBR method, the maximum or minimum bit rate must be limited by resetting the Q scale to a value close to the VBR Q scale in a short cycle. In addition, the Q scale of the CBR method is selected promptly.

As described above, according to the first embodiment of the present invention, the following effects can be achieved with respect to the limitation on the maximum bit rate.
(1) By limiting the maximum bit rate and the minimum bit rate of the VBR method, the image quality deterioration at the change point of the encoding difficulty level and the bit rate is subsequently lowered in order to cancel the excessive bit rate. Image quality degradation can be suppressed.
(2) By determining the remaining amount of GOP using the actual amount of generated bits, the maximum bit rate restriction does not work correctly as the GOP becomes behind, and the amount of generated bits of P and B pictures increases in the latter half of the GOP. The ratio between the I, P, and B pictures of the parameter (global complexity measure) indicating the complexity of the screen used in step 1 collapses, and the target bit amount for limiting the maximum bit rate of the I picture in the next GOP is kept small. Therefore, it is possible to prevent a problem that the image quality of the I picture is deteriorated.
Also, according to the first embodiment of the present invention, the same effect can be obtained on the limitation of the minimum bit rate.
(3) The virtual buffer for obtaining the Q scale for limiting the minimum bit rate is not increased indefinitely, and the Q scale can be quickly lowered when the coding difficulty level changes to an easy picture. , Image quality deterioration during this period can be prevented.

As described above, according to the image coding apparatus and method of the first embodiment of the present invention, it is possible to realize compression coding processing of an image signal at a variable coding rate by a single coding processing.
Furthermore, by utilizing a general encoding processing circuit, there is an advantage that variable encoding rate processing can be realized in applications that require real-time encoding processing without greatly changing the circuit configuration.

Second Embodiment With reference to FIGS. 15 to 18, an image coding apparatus and an image coding method according to a second embodiment of the present invention will be described.
The encoder 20A illustrated in FIG. 15 includes a first VBR buffer recording unit 21-1, a first VBR control unit 22-1, a second VBR buffer recording unit 21-2, a second VBR control unit 22-2, a maximum CBR control unit 23A, and a minimum. It has a CBR control unit 23B, an image compression coding unit 24, a first quantization (Q) scale comparison unit 25A, and a second quantization (Q) scale comparison unit 25B.
The encoder 20A in the image coding apparatus according to the second embodiment of the present invention, the encoder 20 of the first embodiment illustrated in FIG. 6, and the control circuit 10A of the second embodiment and the control of the first embodiment. Differences from the circuit 10 will be described below.

(1) Encoder 20A
The first embodiment and the second embodiment of the present invention are basically the same as the variable coding bit rate (VBR) control circuit 22a illustrated in FIG. 5, but will be described with reference to FIG. The VBR buffer recording unit 21 and the VBR control unit 22 in the first embodiment are different from the first VBR buffer recording unit 21-1 and the first VBR control unit 22- in the second embodiment as illustrated in FIG. 1 and the second VBR buffer recording unit 21-2 and the second VBR control unit 22-2 are provided in parallel in two systems.
(2) Control circuit 10A
The control circuit 10A illustrated in FIG. 15 applies the first target bit rate BR _obj 1 and the first VBR response speed VBRr1 to the first VBR control unit 22-1, and sets the first VBR buffer initial value to the first VBR buffer recording unit 21-1. Apply BR _INI 1 Similarly, the control circuit 10A applies the second target bit rate BR _obj 2 and the second VBR response speed VBRr2 to the second VBR control unit 22-2, and supplies the second VBR buffer initial value BR _INI to the second VBR buffer recording unit 21-2. 2 is applied.
(3) First and second variable bit rate (VBR) quantization scales VBR _q 1 and VBR _q 2 are output from the first VBR control unit 22-1 and the second VBR control unit 22-2, respectively.
(4) The quantization scale selection circuit 25a illustrated in FIG. 5 is the quantization (Q) scale comparison unit 25 in the first embodiment described with reference to FIG. 6, but in the second embodiment, Is provided with a second quantization (Q) scale comparison unit 25B in addition to the first quantization (Q) scale comparison unit 25A corresponding to the quantization (Q) scale comparison unit 25 described with reference to FIG. ing. The processing contents of the first and second quantization (Q) scale comparison units 25A and 25B are shown in the flowchart of FIG. The second quantization (Q) scale comparison unit 25B includes a first variable bit rate quantization scale VBR _q 1 output from the first VBR control unit 22-1, and a second output from the second VBR control unit 22-2. The final variable bit rate quantization scale VBR _q is selected by referring to the variable bit rate quantization scale VBR _q 2 and the quantization threshold Q _th . The first quantization (Q) scale comparison unit 25A performs the same quantization scale QSC selection process as the process of the quantization (Q) scale comparison unit 25 illustrated in FIG.

Other configurations and operations of the image coding apparatus according to the second embodiment are the same as those of the first embodiment. It is described for reference.
(1) Image Encoding Device The configuration of the image encoding device according to the second embodiment of the present invention is the same as that of the image encoding device illustrated in FIG. The image encoding device 1 according to the second embodiment of the present invention includes a control circuit 10A described above instead of the control circuit 10 illustrated in FIG. 4, an encoder 20A described above instead of the encoder 20 illustrated in FIG. 4, and an image input device. 30 is included. The image input device 30 is the same. The basic configuration and operation of the control circuit 10A and encoder 20A are the same as those of the control circuit 10 and encoder 20 described as the first embodiment.

(2) Encoder The basic configuration of the encoder 20A according to the second embodiment of the present invention is also the same as that of the encoder 20 illustrated in FIG. The encoder 20A includes a variable coding bit rate (VBR) control circuit 22a, a fixed coding bit rate (CBR) control circuit 23a, a quantization (Q) scale selection circuit 25a, and an image compression coding unit 24.

Similarly to the first embodiment, the CBR control circuit 23a also uses a first image processing unit, for example, a second image processing unit shorter than a GOP unit, for example, a picture or macro block (MB) unit. The second and third quantization scales QS2 and QS3 that change rapidly so as not to exceed the maximum bit rate specified by the first sequence header of the bitstream Third quantization scale setting means.
Therefore, the fixed bit rate control circuit 23a includes a maximum fixed bit rate (CBR) control unit 23A and a minimum CBR control unit 23B as illustrated in FIG. The configuration of the CBR control circuit 23a illustrated in FIG. 15 is the same as that of the CBR control circuit 23a in the first embodiment described with reference to FIG. The relationship between the configuration of the maximum CBR control unit 23A and the minimum CBR control unit 23B and the image compression encoding unit 24 is the same as that described with reference to FIG.

In the second embodiment, the variable coding bit rate (VBR) control circuit 22a includes the first VBR control unit 22-1 and the first VBR buffer recording unit 21-1, and the second VBR control unit 22-2 and the second VBR. The buffer recording unit 21-2 is configured to operate in two systems in parallel.
The basic operations of the first VBR control unit 22-1 and the second VBR control unit 22-2 are the same as those of the VBR control unit 22 of the first embodiment. That is, the first VBR control unit 22-1 and the second VBR control unit 22-2 perform the first quantization scale (first QQ) for performing image encoding in a first image processing unit, for example, a GOP unit including a plurality of pictures. Is a first quantization scale setting means for setting a first variable bit rate quantization scale VBR _q 1 and a second variable bit rate quantization scale VBR _q 2 as a scale), and a first target bit rate BR _obj 1 and a first variable bit rate quantization scale VBR _q 1 FIG. 16 is a flowchart according to the degree of difficulty of encoding the input image signal (image data S30) to be encoded from the image input device 30 in the first image processing unit with respect to the two target bit rates BR _obj 2. The first quantization scale QS1 (variable bit rate quantization scale VBR _q ) is set by the method shown.

  In the second embodiment, the quantization scale selection circuit 25a is set by the first quantization scale QS1 and the CBR control circuit 23a set by the VBR control circuit 22a of FIG. 5 as in the first embodiment. The second and third quantization scales QS2 and QS3 are compared, and an appropriate Q scale of these quantization scales is encoded with respect to the input image signal (image data S30) to be encoded. The quantization scale QSC is output to the image compression encoding unit 24.

Similarly to the first embodiment, the image compression encoding unit 24 has the configuration illustrated in FIG. 7, and the image data S30 to be encoded is compared with the first quantization (Q) scale. Encoding is performed based on the quantization scale QSC from the quantization scale selection circuit 25a including the unit 25A and the second quantization (Q) scale comparison unit 25B and the compression encoding parameter S12 from the control circuit 10A.
In accordance with the target bit rate BR _obj , the maximum bit rate BR _max , the VBR instruction signal VBD, and the quantization scale QSC provided from the first quantization (Q) scale comparison unit 25A, the image compression encoding unit 24 The image data S30 to be encoded input from the image input device 30 is compression-encoded, and a bit stream S20 is output.
Further, the image compression encoding unit 24 generates a generated bit amount BQT indicating the bit amount generated by the compression encoding and a PCT signal indicating the image encoding type, and as illustrated in FIG. 15, the VBR control unit 22, To the maximum CBR control unit 23A and the minimum CBR control unit 23B.
The image compression encoding unit 24 has the configuration illustrated in FIG. 7, and the operation thereof is the same as that of the first embodiment. Also in the second embodiment, the image compression encoding unit 24 can realize encoding of an image signal at a variable encoding rate by a single encoding process, that is, a one-pass process.

VBR Buffer Recording Unit The operations of the first VBR buffer recording unit 21-1 and the second VBR buffer recording unit 21-2 are the same as those of the VBR buffer recording unit 21 of the first embodiment.
Similarly to the VBR buffer recording unit 21 of the first embodiment, the first VBR buffer recording unit 21-1 and the second VBR buffer recording unit 21-2 are respectively the first VBR buffer recording unit 21-1 and the second VBR buffer recording unit 21-2. The first VBR buffer value VBRb1 and the second VBR buffer value VBRb2 output from the first VBR buffer value VBRb2 are stored, and the total values in the series of encoding processes in the image compression encoding unit 24 described above are the first target bit rate BR _obj 1 And the second target bit rate BR _obj 2 is provided to reduce the deviation amount.
Before starting the encoding process, the first VBR buffer recording unit 21-1 and the second VBR buffer recording unit 21-2 store the first VBR buffer initial value BR _INI 1 and the second VBR buffer initial value BR _INI 2 input from the control circuit 10A. Remember. As the simplest initialization method, the first VBR buffer recording unit 21-1 and the second VBR buffer recording unit 21-2 are initialized according to the initialization control signal from the control circuit 10 A before the processing is started, and the first VBR buffer is initialized. The initial value BR _INI 1 and the second VBR buffer initial value BR _INI 2 are set to “0” or set to predetermined values.
The first VBR buffer recording unit 21-1 and the second VBR buffer recording unit 21-2 are the first VBR buffer recorded in the first VBR control unit 22-1 and the second VBR control unit 22-2 after the encoder 20A starts compression encoding. The initial value BR _INI 1 and the second VBR buffer initial value BR _INI 2 are provided, and the first VBR control unit 22-1 and the second VBR control unit 22-2 receive the first VBR buffer value VBRb1 and the second VBR buffer initial value BR _INI 2 at the end of encoding in the encoder 20A. The 2VBR buffer value VBRb2 is input and recorded. In the next encoding process, the first VBR buffer recording unit 21-1 and the second VBR buffer recording unit 21-2 use the recorded first VBR buffer value VBRb1 and second VBR buffer value VBRb2 as the first VBR control unit 22-1 and second VBR control. Output to the unit 22-2.

VBR Control Unit The operations of the first VBR control unit 22-1 and the second VBR control unit 22-2 are the same as those of the VBR control unit 22 of the first embodiment.
Hereinafter, the first VBR control unit 22-1 will be described. The second VBR control unit 22-2 also sets the first target bit rate BR _obj 1 to the second target bit rate BR _obj 2, the first VBR buffer value VBRb1 to the second VBR buffer value VBRb2, and the first VBR quantization scale VBR _q 1 to the second VBR. only replace the quantizer scale VBR _q 2, is basically the same as the processing of the 1VBR control unit 22-1.

The first VBR control unit 22-1 includes the first variable bit rate according to the first VBR response speed VBRr1, the first target bit rate BR _obj 1, the maximum bit rate BR _max , and the minimum bit rate BR _min input from the control circuit 10A. The (VBR) quantization scale VBR _q 1 is obtained and output to the second quantization (Q) scale comparison unit 25B. The 1VBR control unit 22-1 with respect to the target amount of bits generated 1GOP obtained from the first target bit rate BR _obj 1, comparing actual bit amount generated in the GOP, a first target bit rate BR _obj 1 If the bit amount is excessively generated, the Q scale is gradually increased to gradually suppress the generated bit amount. Conversely, if the bit amount is not generated from the first target bit rate BR _obj 1, the Q scale is gradually increased. Decreasing the amount of generated bits gradually. As a result, the Q scale does not fluctuate with a good response, so the VBR rate control becomes a generated bit amount corresponding to the encoding difficulty level, and the Q scale is gradually changed according to the generated bit amount. The generated bit amount corresponds to the target bit rate BR _OBJ .
In the first VBR control unit 22-1, as the internal parameters, the first remaining in the GOP when encoding with the first VBR buffer value VBRb1 for obtaining the first VBR quantization scale VBR _q 1 and the first target bit rate BR _obj 1 bit amount RBG1, holding the first 1VBR response speed VBRr1 representing the first 1VBR response speed of the quantization scale VBR _q 1.

In the first VBR control unit 22-1, the first VBR quantization scale VBR _q 1 is obtained using the following formula (1a) as in the formula (1).

VBRq1 = VBRb1 / (2 ⁿ × VBRr1) (1a)

When the 1VBR quantizer scale VBR _q 1 is 1, since the VBR rate control becomes unstable bit amount generated is increased rapidly, the 1VBR quantizer scale VBR _q 1 calculated by the equation (1a) is 1 In the following cases, the first VBR control unit 22-1 takes measures such as forcibly changing the first VBR quantization scale VBR _q 1 to 2.
Paying attention to the difference between I / P / B pictures in GOP, the first VBR control unit 22-1 uses a coefficient, for example, 1.4, in the value obtained by equation (1a) when the encoding type is B picture. A value obtained by multiplying is set as the first VBR quantization scale VBR _q 1. That is set the first 1VBR quantization scale VBR _q 1 in the case of B pictures always I, to the 1VBR 1.4 times the quantization scale VBR _q 1 P-picture. The reason for this is that, as described in the first embodiment, the encoding amount that can be saved in the B picture can be reduced by encoding the B picture somewhat coarsely compared to the I and P pictures. , The S / N of the I and P pictures is improved and the S / N of the B picture generated with reference to this is also improved. This is because improvement can be achieved.
In the first VBR buffer recording unit 21-1, the Q scale of the I / P / B picture is determined by a simple method as described above. However, the first VBR buffer recording unit 21-1 is not limited to the above example, and the encoding difficulty level of each picture is determined. It is also possible to add a difference in Q scale according to the examination.

  In the first VBR buffer recording unit 21-1, a target bit amount is set for each GOP as the remaining bit amount RBG in the GOP. In the first VBR buffer recording unit 21-1, for example, when the generated bit amount is received for each MB (macroblock), the corresponding amount (target bit amount for each GOP) is subtracted therefrom. Since the deviation from the target bit amount remains in the remaining bit amount RBG in the GOP at the end of each GOP, the first VBR buffer recording unit 21-1 subtracts this (remaining bit amount RBG) from the first VBR buffer value VBRb1. At this time, when the first VBR buffer value VBRb1 becomes negative, the first VBR control unit 22-1 sets the first VBR buffer value VBRb1 to zero. When the first VBR buffer value VBRb1 is likely to overflow, the first VBR control unit 22-1 fixes the first VBR buffer value VBRb1 to the maximum value.

The initial value of the first VBR buffer value VBRb1 at the start of the encoding process is input from the first VBR buffer recording unit 21-1 to the first VBR control unit 22-1. For example, when the first VBR buffer value VBRb1 input from the first VBR buffer recording unit 21-1 is other than 0, the first VBR control unit 22-1 uses the value. When the first VBR buffer value VBRb1 is 0, the same as the equation (2) The value obtained by the following equation (2a) is used.

VBRb1 = m · VBRr1 / (BR _obj 1 × 2 ^k ) (2a)

  The first VBR response speed VBRr1 in the equation (2a) uses a value supplied by the control circuit 10A. For example, when the first VBR response speed VBRr1 is set to a value of about 200, several tens of seconds can be obtained when the first VBR response speed VBRr1 is set to a value of about tens of seconds and 4000. The first VBR buffer value VBRb1 calculated by Expression (2a) at the end of encoding is transmitted to and stored in the first VBR buffer recording unit 21-1.

In the first remaining bit amount RBG1 in the GOP, a value obtained by the following equation (3a) similar to the equation (3) is first set.

RBG1 = BR _obj 1 · N / PCR (3a)

The second VBR control unit 22-2 also performs the same operation (processing) as the first VBR control unit 22-1 in parallel with the first VBR control unit 22-1.
Also by the second VBR control unit 22-2, since the Q scale does not fluctuate so much in response, the VBR rate control is performed with the generated bit amount corresponding to the encoding difficulty level, and the Q scale is gradually changed according to the generated bit amount. By changing the amount, the generated bit amount according to the second target bit rate BR _obj 2 is obtained in the long term.

CBR Control Unit The CBR control unit of the second embodiment includes a maximum CBR control unit 23A and a minimum CBR control unit 23B, and is the same as the first embodiment described with reference to FIG.
In the maximum CBR control unit 23A and the minimum CBR control unit 23B, similarly to the first embodiment, for example, the CBR rate control is performed on the basis of the TM5 algorithm for MPEG2.
The target bit amount of 1 GOP corresponding to the bit rate by the maximum CBR control unit 23A and the minimum CBR control unit 23B according to the second embodiment, as shown by dotted curves CV22 and CV23 in FIGS. R is set to the remaining GOP bit amount, encoding is controlled by the target bit amount R of each frame obtained from this, a bit amount corresponding to the target bit rate BR _OBJ is generated every frame, and the target bit amount of each frame Is subtracted from the remaining GOP bit amount, the GOP remaining bit amount becomes close to zero at the end of the GOP.

Also in the second embodiment, similarly to the first embodiment, the maximum CBR control unit 23A sets and sets the maximum CBR quantization scale CBR _max q according to the maximum bit rate BR _max input from the control circuit 10A. The CBR _max q is output to the first quantization (Q) scale comparison unit 25A.
Also in the second embodiment, processing is performed under calculations corresponding to the equations (4) to (6).
The maximum CBR control unit 23A and the minimum CBR control unit 23B obtain the maximum CBR quantization scale CBR _max q and the minimum CBR quantization scale CBR _min q based on the target bit amount of the GOP calculated as described above. 1 quantized (Q) scale comparison unit 25A.

Quantization Scale Comparison Unit The processing of the first quantization (Q) scale comparison unit 25A and the second quantization (Q) scale comparison unit 25B in the second embodiment is shown in the flowchart of FIG.
The first target bit rate BR _obj 1 and the second target bit rate BR _obj 2 set in the first VBR control unit 22-1 and the second VBR control unit 22-2 from the control circuit 10A are the second target bit rate BR. _obj 2 is a first greater than the target bit rate _{BR obj 1 (BR obj 2>} BR obj 1).
In the second quantization (Q) scale comparison unit 25B, the first VBR quantization scale VBR _q
1 and the quantization threshold value Q _th are compared, and the smaller one is selected (step 11). Next, the second quantization (Q) scale comparison unit 25B compares the selected one with the second VBR quantization scale VBR _q 2 and sets the larger one as the VBR quantization scale VBR _q and the first quantization (Q) scale. Output to the comparison unit 25A.
As described above, when the VBR quantization scale VBR _q is determined, the first quantization (Q) scale comparison unit 25A performs the VBR quantization scale VBR _q in step 1 in the same manner as described with reference to FIG. Are compared with the minimum CBR quantization scale CBR _min q input from the minimum CBR control unit 23B, and the smaller one is selected. In step 2, the selected quantization scale and the maximum CBR input from the maximum CBR control unit 23A are selected. The quantization scale CBR _max q is compared and the larger one is selected, and the selected scale is calculated as the quantization scale QSC and provided to the image compression encoding unit 24.

  The processing in the first Q scale comparison unit 25A is the same as that described with reference to the flowchart of FIG. In step 1, by comparing VBRq and CBRminq and selecting the smaller one, the maximum value of the Q scale is limited so that the Q scale is not too large and the bit rate is not lowered too much. Next, in step 2, the comparison result and CBRmaxq are compared and the larger one is selected, and the minimum value of the Q scale is limited so that the Q scale is not too small and the bit rate is not increased too much. Other processes of the first Q scale comparison unit 25A are the same as those described in the first embodiment.

The meaning of selecting the VBR quantization scale VBR _q from the first VBR quantization scale VBR _q 1 and the second VBR quantization scale VBR _q 2 in the second Q scale comparison unit 25B will be described.
For example, as illustrated in FIG. 18B, the control circuit 10A sets the low bit rate BR as the first target bit rate BR _obj 1 in the first VBR control unit 22-1, and sets the high bit rate BR to the second value. The target bit rate BR _obj 2 is set in the second VBR control unit 22-2.

In FIG. 17, in the first area Z11, an easy image is input as the image data S30 to be encoded, a difficult image is input in the third area Z13, an easy image is input in the fifth area, and the first area and the third area A case where a first transition region Z12 in which an easy image transitions to a difficult image exists between the regions and a fourth transition region Z14 in which a transition to a difficult but easy image exists between the third region and the fifth region is shown.
A curve VBR _q 1 is a quantization scale Q (BR) assuming that the image data S30 to be encoded is encoded with the first VBR quantization scale VBR _q 1, and a curve VBR _q 2 is a second VBR quantization scale VBR _q . encoded by assuming the quantization scale Q (BR), Q _th denotes the quantization scale Q (BR) when encoding the quantization threshold Q _th fixed. The solid line VBR _q is the quantization scale Q when assuming the encoding process using the quantization scale QSC selected by the first quantization (Q) scale comparison unit 25A and the second quantization (Q) scale comparison unit 25B. (BR) is shown.

Since the first region Z11 is an easy image, the bit rate is controlled based on the first target bit rate BR _obj 1 (BR _obj 2> BR _obj 1), and the first VBR quantization scale VBR _q 1 is quantized. less than threshold Q _th, the 1VBR quantizer scale VBR _q 1 is selected in the first 2Q scale comparison section 25B, based on the VBR _q 1 in the image compression encoding section 24, a lower bit rate BR (high Q scale ) To perform the encoding process.
When the image becomes difficult and reaches the first transition time TP1, VBR _q 1 becomes larger than the quantization threshold Q _th , Q _th is selected by the Q scale comparison unit 25B, and the image compression encoding unit 24 Based on _Qth , encoding processing is performed at a fixed bit rate BR (fixed Q scale).
In the third region Z13 beyond the second transition time point TP2, since the image is difficult, the bit rate is controlled based on the second target bit rate BR _obj 2 and the second VBR quantization scale VBR _q 2 is set to the quantization threshold value. Qth is larger than Q _th and smaller than the first VBR quantization scale VBR _q 1, VBR _q 2 is selected by the second Q scale comparison unit 25 B, and the higher bit is selected based on VBR _q 2 by the image compression encoding unit 24. The encoding process is performed at a rate BR (lower Q scale).
At the fourth transition time Z14 when the image becomes easier beyond the third transition time TP3, VBR _q 2 becomes smaller than the quantization threshold Q _th , and Q _th is selected by the Q scale comparison unit 25B, and the image The compression encoding unit 24 performs encoding processing at a fixed bit rate BR (fixed Q scale) based on _Qth .
Further in the fifth transition point Z15 the image beyond the fourth transition point TP4 is getting more easy, as in the process in the first region Z11, VBR _q 1 is smaller than the Q _th, VBR in Q-scale comparison section 25B _q 1 is selected, and the image compression encoding unit 24 performs encoding processing at a low bit rate BR (high Q scale) based on VBR _q 1.

FIG. 18A is a diagram showing the complexity of the exemplary image and the change in the bit rate BR according to the first embodiment, and FIG. 18B is the complexity of the exemplary image according to the second embodiment. And a change in the bit rate BR.
In the first embodiment, there is one target bit rate BR _obj , and the bit rate BR illustrated in FIG. 18A converges to the same bit rate BR regardless of the difficulty of image encoding. Yes.
In the second embodiment, a first target bit rate BR _obj 1 and a second target bit rate BR _obj 2 are provided, and the bit rate BR illustrated in FIG. For easy images, the bit rate BR is low. For example, assuming that encoded data is recorded on a recording medium, there is an advantage that the recording time on the recording medium can be made longer than only the high bit rate. On the other hand, the bit rate is high for difficult images, and deterioration in image quality can be suppressed compared to only a low bit rate.
Thus, according to the second embodiment, in addition to the features and effects (advantages) of the first embodiment, the amount of encoded data is reduced (based on the degree of difficulty of image encoding) For example, the recording time can be extended), or a predetermined encoded image quality can be maintained.

Modified Embodiment of Second Embodiment The types of target bit rate BR _obj are not limited to the two described above.
For example, as illustrated in FIG. 19, n variable coding bit rate (VBR) control circuits 22a can be arranged in parallel. That is, the first VBR buffer recording unit 21-1, the first VBR control unit 22-1, the second VBR buffer recording unit 21-2, the second VBR control unit 22-2,..., The nth VBR buffer recording unit 21-n, and the second. An nVBR control unit 22-n is provided. The control circuit 10B sets the first VBR buffer initial value BR _INI 1 to the nth VBR buffer initial value BR _INI n in the first VBR buffer recording unit 21-1 to the nth VBR buffer recording unit 21-n, and the first VBR control unit 22 -1 to nth VBR control unit 22-n, the first target bit rate BR _obj 1 to the nth target bit rate BR _obj n, the first VBR response speed VBRr1 to the nth VBR response speed VBRrn, and BR _max and BR _min Set.
Between the first VBR buffer recording unit 21-1 and the first VBR control unit 22-1, and between the nth VBR buffer recording unit 21-n and the nVBR control unit 22-n, the first VBR buffer value VBRb1 to the nth VBR, respectively. Buffer value VBRbn is exchanged.
Of course, the control circuit 10B, which is appropriately set according to the degree of difficulty in encoding image, a first target bit rate BR _obj. 1 to the n-th target bit rate BR _obj n first second 1VBR controller 22-1～ It is set in the nVBR control unit 22-n. As a result, the first VBR quantization scale VBR _q 1 to the nth VBR quantization scale VBR _q n are output from the first VBR control unit 22-1 to the nth VBR control unit 22-n to the second Q scale comparison unit 25C.

FIG. 20 is a flowchart showing an outline of processing of the second Q scale comparison unit 25C.
The 2Q scale comparing unit 25C refers to the plurality of quantization thresholds Q _th 1~Q _th n-1, performed by referring to a similar comparison determination process with Q-scale comparison unit 25B described Figure 16. Thereby, an appropriate quantization scale VBR _q is selected as the variable bit rate.
The first Q scale comparison unit 25A performs a comparison determination process similar to the method illustrated in FIGS.
The bit rate according to the present embodiment is a case where a plurality of target bit rates BR _obj are provided in FIG.

FIG. 1 is a diagram illustrating an example of an image encoding rate by fixed bit rate control. FIG. 2 is a diagram showing a case where the maximum bit rate determined by the standard is exceeded due to the characteristics of the image in the variable bit rate control. FIG. 3 is a diagram showing the quantization scale control result. FIG. 4 is a diagram showing an overall basic configuration of the image encoding apparatus according to the embodiment of the present invention. FIG. 5 is a block diagram showing a basic configuration of the encoder in the present embodiment. FIG. 6 is a block diagram showing a configuration of an encoder as a first embodiment of the encoder illustrated in FIG. FIG. 7 is a block diagram showing a configuration example of the image compression encoding unit in the encoder illustrated in FIG. FIG. 8 is a block diagram showing a configuration example of the maximum CBR control unit and the minimum CBR control unit in the encoder illustrated in FIG. FIG. 9 is a block diagram illustrating a configuration example of the CBR control unit described in Patent Document 1. In FIG. FIGS. 10A to 10C show the states of CBR rate control by the maximum CBR control unit and the minimum CBR control unit illustrated in FIG. 8 shown by thin dashed curves CV21 and CV22, and thick solid curve CV11 to CV13. It is the figure which illustrated the state of the CBR rate control by the CBR control part illustrated in FIG. FIG. 11 is a flowchart showing the processing of the quantization scale comparison unit in the encoder illustrated in FIG. FIGS. 12A and 12B are diagrams illustrating quantization scale control by the image coding apparatus according to the first embodiment of this invention. FIG. 13 is a diagram illustrating an example of a Q scale control result by the image coding apparatus according to the first embodiment of the present invention. FIG. 14 is a flowchart showing the overall operation of the image coding apparatus according to the first embodiment of the present invention. FIG. 15 is a block diagram showing a configuration of an encoder as a second embodiment of the encoder illustrated in FIG. FIG. 16 is a flowchart showing the processing of the quantization scale comparison unit in the encoder illustrated in FIG. FIG. 17 is a diagram illustrating an exemplary change in the quantization (Q) scale by the image coding apparatus according to the second embodiment of the present invention. FIGS. 18A and 18B are diagrams showing exemplary bit rate changes by the image coding apparatuses according to the first and second embodiments of the present invention. FIG. 19 is a block diagram showing a configuration of an encoder as a modification of the second embodiment of the encoder illustrated in FIG. FIG. 20 is a flowchart showing part of the processing of the quantization scale comparison unit in the encoder illustrated in FIG.

Explanation of symbols

1. Image encoding device
10 ... Control circuit, 20 ... Encoder (encoding device), 30 ... Image input device
20 ... Encoder (encoding device)
21 ... VBR buffer recording section
22 ... VBR controller
23A: Maximum CBR controller
2311 ... GOP allocation bit amount calculation unit
2312 ... GOP remaining bit amount buffer
2313 ... TM5 Step 1 operation unit
2314 ... TM5 Step 2 operation unit
23B ... Minimum CBR control unit
2321 ... GOP allocated bit amount calculation unit
2322 ... GOP remaining bit amount buffer
2323 ... TM5 Step 1 operation unit
2324 ... TM5 Step 2 operation unit
24: Compression encoding unit
212 ... Coding control unit
25 ... Quantization scale comparator
25A 1st Q scale comparison part
25B, 25C ... 2nd Q scale comparison part

Claims (4)

Image coding that encodes an image signal to be encoded by a response bit rate, a plurality of target bit rates, a maximum bit rate, and a minimum bit rate, which are given from an external device, using a variable bit rate method and a constant bit rate method. A device,
To achieve the above given above SL plurality of target bit rate when encoding an image signal of the coding target at a variable bit rate system, according to the coding difficulty of the image signal in the first image processing unit First quantization scale calculation means for calculating a plurality of first quantization scales that change
When the image signal to be encoded is encoded by the fixed bit rate method, the second quantum processing unit is smaller than the first image processing unit so as not to exceed the given maximum bit rate. A second quantization scale calculating means for calculating a quantization scale;
When the image signal to be encoded is encoded by a fixed bit rate method, it is smaller than the first image processing unit so as not to fall below the minimum bit rate given according to the image signal to be encoded. Third quantization scale calculating means for calculating a third quantization scale in a third image processing unit;
A larger quantization scale is selected by comparing the first quantization scale and the second quantization scale, and a smaller quantization scale is compared by comparing the selected quantization scale with the third quantization scale. A quantization scale selection means for selecting
Coding means for performing a coding process on the image signal to be coded using the quantization scale selected by the quantization scale selection means;
To the first quantization scale calculating means for calculating said first quantization scale, the given plurality of was in accordance with the coding difficulty when encoding an image signal of the coding target at a variable bit rate system target bit rate and the amount of Coca threshold is set,
The first quantization scale calculation means calculates a plurality of first quantization scales in the first image processing unit in order to achieve each of the set target bit rates.
The quantization scale selection means is at least:
The smaller one by comparing the first quantization thresholds corresponding to the first quantization scale of the first quantization scale and the quantization thresholds of the plurality of first quantization scale first select a value of 1, select the greater by comparing the second quantization scale of the first value and the plurality of first quantization scale and the selected as the second value,
The greater by comparing the second amount coca scale calculated by the second value and the second quantization scale calculation unit described above selects selected as the third value, the third value obtained by the selected selecting smaller as the fourth value is compared with the third quantization scale calculated in the third quantization scale calculation means,
Doing
Outputting the finally selected fourth value to the encoding means;
Image encoding device. The encoding means encodes an input image signal based on the MPEG2 standard using the fourth value finally selected by the quantization scale selection means.
The image encoding device according to claim 1. Image coding that encodes an image signal to be encoded by a response bit rate, a plurality of target bit rates, a maximum bit rate, and a minimum bit rate, which are given from an external device, using a variable bit rate method and a constant bit rate method. A method,
To achieve the above given above SL plurality of target bit rate when encoding an image signal of the coding target at a variable bit rate system, according to the coding difficulty of the image signal in the first image processing unit A first quantization scale calculating step of calculating a plurality of first quantization scales that change
When the image signal to be encoded is encoded by the fixed bit rate method, the second quantum processing unit is smaller than the first image processing unit so as not to exceed the given maximum bit rate. A second quantization scale calculating step for calculating a quantization scale;
When the image signal to be encoded is encoded by a fixed bit rate method, it is smaller than the first image processing unit so as not to fall below the minimum bit rate given according to the image signal to be encoded. A third quantization scale calculating step of calculating a third quantization scale in a third image processing unit;
A larger quantization scale is selected by comparing the first quantization scale and the second quantization scale, and a smaller quantization scale is compared by comparing the selected quantization scale with the third quantization scale. A quantization scale selection step of selecting
An encoding step of performing an encoding process on the image signal to be encoded using the quantization scale selected by the quantization scale selection step;
Oite to the first quantization scale calculation step of calculating the first quantization scale, given above in accordance with the coding difficulty when encoding an image signal of the coding target at a variable bit rate system bit rate and the amount Coca threshold of the plurality of targets are set,
In order to achieve each of the plurality of target bit rates set in the first quantization scale calculation step, a plurality of first quantization scales are calculated in the first image processing unit,
In the quantization scale selection step, at least,
More smaller than the first quantization thresholds corresponding to the first quantization scale of the first quantization scale and the upper SL amount Coca threshold among the plurality of first quantization scale was selected as the first value, it selects the greater by comparing the second quantization scale of the first value and the plurality of first quantization scale and the selected as the second value,
The greater by comparing the second amount coca scale calculated by the second value and the second quantization scale calculation step of the above selected selected as the third value, the third value obtained by the selected selecting smaller as the fourth value is compared with the third amount Coca scale calculated above Symbol third quantization scale calculation step,
Doing
The finally selected fourth value is used in the encoding step.
Image coding method. In the encoding step, the input image signal is encoded based on the MPEG2 standard using the fourth value finally selected in the quantization scale selection step.
The image encoding method according to claim 3.

Images