JPH06113334A - Quantization system for three-dimension volume data - Google Patents

Abstract

PURPOSE:To reduce the information quantity and to improve the coding efficiency by utilizing a visual characteristic so as to change a quantized value depending on this side of picture and a depth of the picture. CONSTITUTION:Quantization sections 1-1 to 1-K are used to quantize data of each hierarchy in three-dimension volume data having a depth hierarchy subjected to coding not including quantization, and after each quantization output signal is variable length coded by a variable length coding section 2, the result is sent to a buffer 3. Only one buffer is used from this side to the depth and the representative quantization value Q by the residual quantity is calculated by a representative quantization value calculation section 4. After the representative quantization value Q is calculated, the representative quantization value Q is used and a quantization value calculation section 5 decreases the quantized picture this side when viewed from the observer and increases the quantization value of the depth picture through the calculation of the quantized values Q1-QK, and the result is given to the quantization values Q1-QK and the result is sent to the receiver side. That is, the conspicuous this side picture is quantized in details to enhance the visual characteristic of the observer.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a quantization system for three-dimensional volume data, and more particularly to a system for quantizing the data of each layer of volume data for three-dimensional stereoscopic image display having depth layers. .

Three-dimensional volume data is a type of three-dimensional stereoscopic image display data, and each point in the three-dimensional space has a value. Depending on how this image data appears,
It can be divided into the following three.

In the case of having not only the surface of the subject but also image data of the contents: The three-dimensional image data obtained by CT scanning or the like has information of the contents for seeing through the contents of the subject, and is suitable for medical use. It is often handled as image data. When there is image data of only the surface of the subject: When image data of the subject is captured from different angles using multiple cameras and depth is estimated. Therefore, there is no image data of the contents of the subject (this is called null data or NULL as described later). When only the image data of the surface when the subject is viewed from a certain direction is included: The image data of the subject is 1 from a certain angle.
When using one or more cameras to estimate the depth. Therefore, there is only the image data of the surface when the subject is viewed from that direction, and naturally, there is no image data of the contents, and it is null data.

As described above, since the three-dimensional volume data has a depth hierarchy, it is expected to be used in various ways as compared with the conventional two-dimensional image data. Therefore, it is necessary to optimize its quantization method. .

[0005]

2. Description of the Related Art FIG. 10 is a block diagram showing a well-known image data encoding system.
First, the screen (space) data is divided into blocks (cubes) and input. The input data is subjected to orthogonal transformation, DPCM, VQ + SQ, or other preprocessing in a coding unit 10 that does not include quantization, and then is quantized in a quantizer (Q) 11. This quantized value becomes a quantization number, and this quantization number is further variable-length coded by the variable-length coding unit (VLC) 12, and a certain time depends on the congestion degree of the transmission path (not shown). Only, the remaining amount is stored in the buffer 13 as the remaining amount of the buffer and transmitted.

Then, the quantized value calculator 14 calculates the quantized value of the quantizer 11 according to the remaining amount of the buffer 13, selects a corresponding quantizer from a plurality of quantizers, and The encrypted value is transmitted to the receiving side.

The relationship between the remaining amount of the buffer 13 and the quantized value will be roughly described. When the bit rate of the transmission path is constant, the remaining amount of the buffer 13 increases as the generated information amount increases. Therefore, in order to prevent the remaining amount of the buffer 13 from increasing, the quantization value is calculated and controlled so as to select a rough quantizer (parameter). On the contrary, if the generated information amount is small, the remaining amount of the buffer 13 is small. Therefore, in order to prevent the remaining amount of the buffer 13 from being out of balance, the quantization value is calculated so that a fine quantizer (parameter) is selected. Take control.

More specifically, the following three known methods are well known as methods for calculating the quantized value.

(1) When the transmission line has a constant bit rate (C
CITT standard H.264. 261): A method of determining a quantized value using the remaining amount of the data output buffer from the transmitting side as a parameter. In the 261 reference model RM8,
The quantized value is determined by the following formula. Quantized value Q = 2 × INT (BUFF / (200 × q))
+2 formula (1) BUFF is the remaining amount of the buffer 13, that is, the buffer 13
, The number of bits remaining in q, q is the transmission line bit rate (q × 64 kbps). This equation represents a feedback control in which when the amount of information is large and the remaining amount BUFF of the buffer 13 increases, the quantization value also increases, and the amount of information generated thereafter can be suppressed low.

(2) When quantizing in consideration of the above method (1) and image properties (ISO standard MPEG2): method (1)
The quantized value is calculated in consideration of visual characteristics (strictly speaking, a modification of method (1)). That is, the variance of blocks is calculated, and the variance of blocks that have already been quantized is also used as a parameter, and the quantization value is increased in the fine pattern portion (the variance value is large) and the flat pattern portion (variance value is large). Is small)
Then, the calculation is performed so that the quantized value is reduced.

(3) When the transmission line has a variable bit rate: A
In the case of TM or the like, packet transmission is performed even if the quantized value is fixed and information is generated in bursts. Therefore, the above method (1),
It is not necessary to control at a constant transmission bit rate as in (2). However, even in this case, an efficient encoding (quantization) method while maintaining the image quality is desired.

[0012]

In the above conventional method, the quantizer is constant in the depth hierarchy direction even if the input data is three-dimensional volume data having depth hierarchy. In other words, the amount of information is important because the quantized values of the space in the foreground and the space in the back, which have significant meaning in visual characteristics (when there are objects in the foreground and in the back of the space, the gaze concentrates on the object in the foreground), are not changed. There is a problem that the encoding efficiency increases and the encoding efficiency is low.

Therefore, it is an object of the present invention to reduce the amount of information and improve the coding efficiency in a method of quantizing the data of each layer of volume data for three-dimensional stereoscopic image display having a depth layer. And

[0014]

In order to achieve the above object, in the three-dimensional volume data quantization method according to the present invention, as shown in principle in FIG. 1, a code that does not include quantization is used. Quantization units 1-1 to 1-K that perform quantization on the data of each layer of the three-dimensional volume data having a depth layer that has been subjected to quantization, and each quantization unit 1-1 to 1-
A variable-length coding unit 2 that collectively variable-length-codes K output signals, a buffer 3 that temporarily stores the output signal of the variable-length coding unit 2 and outputs the output signal according to the state of the transmission path, and the buffer 3 Of the representative quantization value Q from the remaining amount, and from the representative quantization value Q, the quantization value of the image in front of the viewer is reduced to quantize the image in the back. Quantized value calculator 5 is provided for calculating each quantized value Q1 to QK so as to increase the value, giving it to each quantizer 1-1 to 1-K, and transmitting it to the receiving side.

That is, the quantizers 1-1 to 1-K perform quantization on the data of each layer of the three-dimensional volume data having the depth layer encoded without quantization.
The quantized output signals are variable length coded by the variable length coding unit 2 and then sent to the buffer 3.

At this time, since there is only one buffer from the front to the depth, the representative quantized value calculating unit 4 calculates the representative quantized value Q according to the remaining amount. After calculating the representative quantized value Q, the quantized value calculating unit 5 uses the representative quantized value Q to quantize values Q1 to Q from the front to the depth.
K is calculated and given to each of the quantizers 1-1 to 1-K and transmitted to the receiving side.

Further, in the three-dimensional volume data quantization method according to the present invention, as shown in principle in FIG.
3 with a depth layer that has been encoded without quantization
Quantization units 1-1 to 1-K that perform quantization on the data of each layer of the dimensional volume data, and quantization units 1-1 to 1-1.
Variable-length coding units 2-1 to 2-K for variable-length coding the output signals of 1-K, and the variable-length coding units 2-1 to 2-1.
Buffers 3-1 to 3-K for temporarily accumulating each output signal of -K and outputting according to the state of the transmission line, and the buffers 3-1 to 3-1
The quantization value for the corresponding quantization unit 1-1 is calculated as the representative quantization value Q1 from the remaining amount of the buffer 3-1 corresponding to the image closest to the viewer in 3-K. Based on the representative quantizing value calculation unit 40 and the representative quantizing value Q1, another quantizing unit reduces the quantizing value of the image in front of the viewer and increases the quantizing value of the image in the back. 1-
Quantization value calculator 50 which calculates and gives the respective quantized values Q2 to QK for 2 to 1-K and transmits them to the receiving side.

In this case, the quantizer 1 is applied to the data of each layer of the three-dimensional volume data having the depth layer.
Quantization is performed by -1 to 1-K, and each quantized output signal is variable-length coded by the corresponding variable-length coding units 2-1 to 2-K, and then the corresponding buffer 3-1 is also used. ~ 3
-Send to K.

Then, the representative quantized value calculator 40 calculates a representative quantized value Q1 according to the remaining amount of the buffer 3-1 corresponding to the image closest to the viewer, and the corresponding quantized value is calculated. Give to part 1-1.

Further, the quantized value calculation section 50 receives the representative quantized value Q1 and quantized values Q2 to Q from the front to the depth.
K is calculated and given to each quantizer 1-2 to 1-K and transmitted to the receiving side.

As described above, according to the present invention, the quantization value of the image in front of the viewer is small, that is, finely quantized,
By increasing the quantization value of the image in the back, that is, by coarsely quantizing the image, the image in the foreground, which is more conspicuous, can be more finely quantized, and the visual characteristics of appearance are improved.

In the above case, the quantized value calculation unit 5 uses the representative quantized value, the image variance, the number of data for which no value can exist, and the depth of the depth image as parameters in each layer. The quantized value of can be calculated.

Further, only the above-mentioned representative quantized values Q and Q1 are transmitted to the receiving side, and the quantized value calculators 5 and 50 change the quantized values in the depth direction based on the representative quantized values Q and Q1. Each quantization value is calculated by setting the amount constant and each quantization unit 1-1 to
1-K (in the case of the quantized value calculation unit 50, 1-2 to 1-
It is also possible to give it to K) so that it is not transmitted to the receiving side.

Further, the quantized value calculators 5 and 50 allow the reception side to share the variation amount of the quantized value in the depth direction without transmitting it, so that the past buffer remaining amount and the past quantized value in each layer can be shared. Can be used as a parameter to determine the quantized value in each layer.

Further, when the transmission line has a variable bit rate,
The buffer 3 (3-1, 3-k) becomes an unnecessary non-feedback type, and the representative quantized value calculation units 4 and 40 do not transmit the representative quantized value Q as a constant and the representative quantized value Q It is also possible for the quantized value calculation units 5 and 50 to calculate and transmit each quantized value based on the above.

[0026]

Embodiment (No. 1) : FIG. 3 shows an embodiment of a representative quantized value calculation unit 4 and a quantized value calculation unit 5 in the quantization method of three-dimensional volume data according to the present invention shown in FIG. 3 shows a calculation algorithm, and FIG. 4 shows hardware for realizing the algorithm shown in FIG.
An embodiment of the representative quantized value calculation and the quantized value calculation will be described with reference to FIG. In this embodiment, only the representative quantized value Q is transmitted, and the amount of change in the quantized value of the image in each depth direction is constant.

First, the value of the representative quantized value Q is calculated with the remaining amount of the buffer 3 as B, and the change ratio with respect to the representative quantized value Q of each depth layer is set to Xn = X1, ..., XK (FIG. 3 step S1).

The representative quantized value Q is calculated by the above-mentioned equation (1).
CCITT standard H. The quantized value calculation method of H.261 is adopted (step S2).

Then, for example, when the number of blocks in the depth direction is "8" in a certain stereoscopic image, X1 =
0.7, X2 = 0.8, X3 = 0.9, X4 = 1.0,
X5 = 1.0, X6 = 1.1, X7 = 1.2, X8 =
The value of X is set to 1.3 so that the average value of X becomes 1 (step S3).

The quantized value Qn of the image in each depth direction
= Q × Xn (n = 1 to K, and the same applies hereinafter) (step S4), and this circuit block diagram is shown in FIG. This is a simple one in which each representative quantized value Q is multiplied by a constant rate of change Xn using a multiplier to calculate each Qn.

In the embodiment shown in FIGS. 3 and 4, the representative quantized value calculator 40 and the quantized value calculator 50 shown in FIG. 2 are used.
Also in the above, there is only a difference between using the foremost quantized value Q1 and using the representative quantized value Q, and the others are common.

Embodiment ( No. 2) : FIG. 5 shows another embodiment of the representative quantized value calculation unit 4 and the quantized value calculation unit 5 in the quantization system for three-dimensional volume data according to the present invention shown in FIG. 6 shows the calculation algorithm in FIG. 6, and FIGS. 6 to 7 show the hardware that realizes the algorithm in FIG. 5. Hereinafter, the representative quantized value calculation and the quantized value calculation will be described with reference to FIGS. Will be described. In this embodiment, the representative quantized value is used only as a parameter on the transmitting side, and the quantized values Q1 to QK of each depth are transmitted. Therefore, the amount of change in the quantized value at each depth varies depending on the nature of the image.

First, the buffer remaining amount is B, the representative quantized value is Q, the average of the variance of the cube (block) of the foreground screen is V1, and null data (pixels for which there is no information such as the contents of the object or the back side) ) Is N1, ..., The average of the variances of the cubes (blocks) in the innermost screen is Vn, and the number of null data is N
n and the number of layers is K (step S11).

Thereafter, the representative quantized value Q is calculated in the same manner as in step 2 of FIG. 3 (step S12).

Individual calculated values Vn of the above-mentioned variances V1 to VK
Is, as is known, and as shown in FIG. 6, a plane pixel array C for each layer output from the scan converter 20.
Square error (Cn (i) -Cn ') ² between n (i) and its average value Cn'
Is an average value obtained by dividing the cumulative value of by the number of calculated pixels that are not null data (step S13). Since this null data part has no information, it cannot be included in the variance calculation.

Next, the value of the evaluation function NUN is calculated. The meaning of this NUN is that the smaller the variance Vn is, the flatter the image of the block becomes, and the smaller NUN is, and the larger the number Nn of null data is. The number of pixels to be transmitted is small (even if it is transmitted, it is regarded as a relatively flat portion).
Is to show that it becomes smaller.

Therefore, the evaluation function NUn is (Vn + 1)
N '/ (Nn + 1) (N' is the average value of the null data Nn) (step S13). The larger the evaluation function is, the coarser the quantizer is. The smaller the quantizer is, the smaller the quantizer is. It should be set finely (quantization value is small).

Further, a second evaluation function Xn is determined as an average value U'of all evaluation functions Un (step S13).
And, although the prototype of this equation is known, the representative quantized value Q is actually
Is used for the purpose of continuously changing 1 / a to a times.

How to obtain the parameter a is shown in FIG.
The larger the depth, the larger it becomes. That is, assuming that the number of blocks in the depth direction is K and the current depth is the nth block, a _n = 1 + 2 × n / K.

The evaluation function Xn relating to the final rate of change is obtained by using this a. When the depth n is K / 2 or less and Xn is larger than 1 for the hierarchy on the front side of half (step S14), X = 1 is set (step S15) and the quantized value should be coarse in calculation, but since it is near the front screen, the rate of change Xn is forcibly made fine to "1". Therefore, the quantized value is "1" or less in the front half layer.

If the depth n is K / 2 or more and X is smaller than 1 (step S16), X = 1 is set and a limiter is applied (step S17). This is because the quantized value should be finely calculated, but since it is far from the screen in front, the rate of change Xn is forcibly made coarse to "1". Therefore, the quantized value is "1" in the latter half layer.
That is all.

Incidentally, these steps S14 to S17 are
This is performed by the limiters (LIM) 21-1 to 21-K shown in FIG.

Then, the quantized values Q1 to QK are obtained as representative values Q × X1 to Q × XK (step S18).

In this way, the quantized value calculator 5
The quantized value in each layer is adaptively calculated using the representative quantized value, the variance of the image, the number of data in which no value can exist, and the depth of the depth image as parameters.

The representative quantized value calculator 40 and the quantized value calculator 50 shown in FIG. 2 are also used in the embodiments shown in FIGS. 6 and 7.
In the above, there is only a difference between using the foremost quantized value Q1 and using the representative quantized value Q, and the others are common.

Embodiment ( No. 3) : FIG. 8 is another embodiment of the representative quantized value calculation unit 4 and the quantized value calculation unit 5 in the quantization system for three-dimensional volume data according to the present invention shown in FIG. 9 shows the calculation algorithm in FIG. 9, and FIG. 9 shows the hardware that realizes the algorithm in FIG. 8. Hereinafter, an embodiment of the representative quantized value calculation and the quantized value calculation will be described with reference to FIGS. . In this embodiment, only the representative quantized value Q is transmitted, and the amount of change in the quantized value for each depth varies.

First, it is assumed that the representative quantized value is Q, and the past (one frame before) information amount of each depth (equivalent to the buffer remaining amount) is Bn = B1 to BK and is counted by the counter 30. Also, the past encoded quantized value is Qn = Q1
~ QK. Such past buffer remaining amount and quantized value may be stored in a memory (not shown), and can be counted by the receiving side. Further, in order to continuously change the representative quantized value Q to 1 / a to a times, the parameter a is set to a _n = 1 + 2 × n / K (step S).
21).

Thereafter, the representative quantized value Q is calculated in the same manner as in step 2 of FIG. 3 (step S22).

Next, let B'be the average value of B1 to BK, and Q
The average value of 1 to QK is set to Q '(step S23).

Next, the evaluation function Un = (Bn / B ') ×
(Qn / Q ') is calculated. This evaluation function Un is larger than the average value of the generated information amount, and the representative quantized value Q
It means that the larger the past encoded quantized value is, the larger the present quantized value should be larger than the representative quantized value, which is a parameter indicating the ease of encoding. Then, the average value U'of this evaluation function Un is obtained (step S24).

A final evaluation function Xn is obtained using the above evaluation function Un, the average value U'and the parameter a in the same manner as in the embodiment of FIG. 5 (step S25).

Then, when the depth n is K / 2 or less and Xn is 1
If it is larger (step S26), X = 1 is set (step S27) to force the change rate Xn to "1", and if the depth n is K / 2 or more and X is smaller than 1 (step S28). ), A limiter is applied with X = 1 (step S27). This is performed by the limiters (LIM) 40-1 to 40-K of FIG. 6 as in the case of FIG.

Then, the quantized values Q1 to QK are obtained as representative values Q × X1 to Q × XK (step S29).

8 and 9, the representative quantized value calculating section 40 and the quantized value calculating section 50 in FIG. 2 either use the frontmost quantized value Q1 or the representative quantized value. The only difference is whether to use Q, and the others are common.

In each of the above embodiments, when the transmission line has a variable bit rate, the buffer 3 (3-1 to 3-)
k) is not used, the representative quantized value calculation unit 4 does not transmit the representative quantized values Q, Q1 as constant, and the representative quantized values Q, Q1 are not transmitted.
The quantized value calculator 5 calculates each quantized value based on Q1 and transmits it.

[0056]

As described above, according to the quantization method for three-dimensional volume data according to the present invention, a representative quantized value is obtained from the remaining buffer capacity, and the representative quantized value is used as a foreground for the viewer. Each quantization value is calculated and given to each quantization unit so as to decrease the quantization value of the image and increase the quantization value of the back image, and the representative quantization value and / or the quantization value is given to the receiving side. Since it was configured to transmit, by using the visual characteristics, by changing the quantization value in the front and back of the image, it is possible to finely quantize the image of the front that is conspicuous,
It is possible to improve the appearance characteristics.

[Brief description of drawings]

FIG. 1 is a block diagram showing a principle configuration (No. 1) of a three-dimensional volume data quantization method according to the present invention.

FIG. 2 is a block diagram showing a principle configuration (No. 2) of the three-dimensional volume data quantization method according to the present invention.

FIG. 3 is a flow chart diagram showing an embodiment (part 1) of the quantization system for three-dimensional volume data according to the present invention.

FIG. 4 is a circuit block diagram of an embodiment (1) of the present invention.

FIG. 5 is a flowchart showing an embodiment (No. 2) of the quantization system for three-dimensional volume data according to the present invention.

FIG. 6 is a circuit block diagram (No. 1) of the embodiment (No. 2) of the present invention.

FIG. 7 is a circuit block diagram (No. 2) of the embodiment (No. 2) of the present invention.

FIG. 8 is a flowchart of an embodiment (part 3) of the quantization system for three-dimensional volume data according to the present invention.

FIG. 9 is a circuit block diagram of an embodiment (3) of the present invention.

FIG. 10 is a block diagram showing a conventional method.

[Explanation of symbols]

 1-1 to 1-K Quantization unit 2, 2-1 to 2-K Variable length coding unit 3, 3-1 to 3-K buffer 4,40 Representative quantized value calculation unit 5,50 Quantized value calculation In the drawings, the same reference numerals indicate the same or corresponding parts.

Front Page Continuation (72) Inventor Kiichi Matsuda 1015 Kamiodanaka, Nakahara-ku, Kawasaki-shi, Kanagawa Fujitsu Limited

Claims (6)

[Claims] 1. A quantizer (1-1 to 1-K) for quantizing data of each layer of three-dimensional volume data having a depth layer encoded without quantization, and A variable length coding unit (3) for collectively variable length coding the output signals of the quantizing units (1-1 to 1-K), and temporarily storing and transmitting the output signal of the variable length coding unit (3) A buffer (3) that outputs according to the state of the road, a representative quantization value calculation unit (4) that calculates a representative quantization value (Q) from the remaining amount of the buffer (3), and a representative quantization value ( From Q), each quantization value (Q1 to QK) is calculated so that the quantization value of the image in the foreground is small and the quantization value of the image in the back is increased from the viewpoint of the viewer.
A quantization method for three-dimensional volume data, comprising: a quantization value calculation unit (5) which is given to (1-1 to 1-K) and is transmitted to the reception side. 2. Quantization units (1-1 to 1-K) for quantizing data of each layer of three-dimensional volume data having a depth layer encoded without quantization, and A variable length coding unit (2-1 to 2-K) for variable length coding the output signals of the quantizing units (1-1 to 1-K), respectively, and a variable length coding unit (2-1 to 2). -K) buffers (3-1 to 3-K) that temporarily store each output signal and output it according to the state of the transmission line, and a viewer of the buffers (3-1 to 3-K) The representative quantization value obtained by calculating the quantization value for the corresponding quantization unit (1-1) from the remaining amount of the buffer (3-1) corresponding to the frontmost image as the representative quantization value (Q1). By the value calculation unit (40) and the representative quantization value (Q1), another quantization unit is used so as to decrease the quantization value of the image in front of the viewer and increase the quantization value of the image in the back. Each quantized value for (1-2 to 1-K)
A three-dimensional volume data quantizing method, comprising: a quantizing value calculating unit (50) for calculating and giving (Q2 to QK) and transmitting it to the receiving side. 3. The quantized value calculation unit (5, 50) uses the representative quantized value, the image variance, the number of data in which no value can exist, and the depth of the depth image as parameters. The three-dimensional volume data quantization method according to claim 1 or 2, wherein the quantization value in (3) is calculated. 4. Only the representative quantized value (Q, Q1) is transmitted to the receiving side, and the quantized value calculation unit (5, 50) determines the depth direction based on the representative quantized value (Q, Q1). Each quantizer (1-1 ~ 1-K)
And not transmitting to the receiving side.
Alternatively, the three-dimensional volume data quantization method described in 2. 5. The past buffer remaining amount and the past buffer remaining in each layer so that the quantization value calculation unit (5, 50) can share the variation amount of the quantization value in the depth direction without transmitting it. 3. The three-dimensional volume data quantization method according to claim 1, wherein the quantization value in each layer is determined using the quantization value as a parameter. 6. When the transmission path has a variable bit rate, the buffer (3) is not used and the representative quantization value calculation unit (4) does not transmit the representative quantization value (Q, Q1) as a constant value. The quantized value calculating unit (5) calculates and transmits each quantized value based on the representative quantized value (Q, Q1).
Alternatively, the three-dimensional volume data quantization method described in 2.