JP2002152776A - Method and device for encoding and decoding distance image - Google Patents

Abstract

PROBLEM TO BE SOLVED: To encode a distance image with high encoding efficiency while holding practical distance information by utilizing a difference in the measuring resolution of the distance image for each point on the surface of an object. SOLUTION: On the side of distance image encoding, after the distance image having the information of a distance which is measured by a distance image sensor, to each point on the surface of the object as each of pixel values is sampled, quantizing width is set for each portion of the distance image on the basis of the resolution of the distance image sensor corresponding to the relevant portion, and the distance image is quantized and encoded. For example, the quantizing width of each portion is set to the distance between the portion and the sensor in proportion to a square of the distance. On the side of distance image decoding, encoded data are decoded and inversely quantized on the basis of the quantizing width of each portion and the distance image is restored.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an image encoding and decoding technique, and more particularly to an encoding technique for compressing a distance image at a high rate of friction and its decoding technique.

[0002]

2. Description of the Related Art To record and transmit the appearance of a three-dimensional world, two-dimensional information of the color, luminance, and distance of a target surface from a certain viewpoint, that is, a color image or a luminance image and a color difference image, and a camera. A range image having an absolute distance to each point on the object surface as a value of each pixel is required.
To store and transmit information of a more complete three-dimensional world, a color image and a range image from multiple viewpoints are required. Further, in order to reproduce the appearance from an arbitrary viewpoint on the reproducing side, it is necessary to perform arithmetic processing on information from more viewpoints and reintegrate the information. For that purpose, it is necessary to efficiently compress a large number of color images and range images.

Generally, an analog range image input from a range image sensor is converted into a discrete-time signal by sampling, and is converted into a discrete-value signal by quantization. In the prior art, linear quantization has been performed in which the entire amplitude axis of the range image signal is uniformly quantized with a uniform width.

On the other hand, as a technique for efficiently encoding a distance image, a method for predictive encoding of a distance image can be considered.
The predictive coding method focuses on the property that the correlation between adjacent pixels is high, predicts the current pixel value from the immediately coded pixel value, and codes the residual signal.
This is also used for encoding a distance image. Further, as a method of efficiently encoding a distance image, the present applicant disclosed in Japanese Patent Application No. 2000-239913 a pixel of a distance image using a correlation between a distance image and a luminance image.
A method of reversibly encoding a distance image with high efficiency by using a model that predicts from a pixel of a luminance image corresponding to an adjacent pixel of the distance image, and a method of decoding the encoded distance image Has been proposed. But,
The quantization is not specifically described.

[0005]

It is known that the measurement resolution of a range image differs at each point on the surface of an object. Here, the measurement resolution indicates the ability to accurately measure the distance from the sensor to the object surface. When the resolution is high, the measurement error (the difference between the true value and the measured value) of the distance from the sensor to the object surface is small. Conversely, when the resolution is low, the measurement error of the distance from the sensor to the object surface is large. In a normal range image measurement, data is stored with the same number of bits (for example, 16 bits for each pixel) regardless of whether the resolution is high or low. This means that a substantially large number of bits are assigned to the distance information of the portion where the measurement resolution is low. Therefore, in the conventional linear quantization, there is a limit in compression of a range image.

An object of the present invention is to take advantage of the fact that the measurement resolution of a range image is different at each point on the object surface, and to compress the range image with high efficiency while visually maintaining almost the same estimation accuracy as lossless coding. It is an object of the present invention to provide an encoding method and an apparatus for performing the above-mentioned encoding, and a method and an apparatus for decoding the encoded distance image.

[0007]

According to the present invention, when encoding a range image, the range image is measured for each part by utilizing the fact that the measurement resolution of the range image is different at each point on the object surface. The main feature is that the quantization width is set based on the resolution of the range image sensor corresponding to the portion. That is, the quantization width of the high-resolution part is set to be small and the quantization width of the low-resolution part is set to be large so that the quantization width of each part monotonously decreases with the height of the resolution. In a specific example, using the fact that the resolution of the range image sensor is proportional to the square of the distance from the sensor, the quantization width of each part of the range image is determined by the distance between the relevant part of the object surface and the range image sensor. Is set in proportion to the square. As a result, the amount of data to be stored can be deleted while retaining substantial distance information, and encoding of a distance image with excellent encoding efficiency is realized.

Hereinafter, the measurement resolution of a distance image used in the present invention will be described. Methods for obtaining the three-dimensional shape of an object in the real world are roughly classified into a “passive method” and an “active method”. The “passive method” is a method of measuring a three-dimensional shape of an object without projecting any meaningful energy to the object, and includes stereo vision and the like. On the other hand, the “active method” obtains a three-dimensional shape by projecting light, sound waves, and the like on a measurement target object and measuring the behavior. Generally, both the “passive method” and the “active method” measure and record the distance from the camera to each point on the object surface based on the principle of triangulation.

A sensor that obtains the three-dimensional shape of a target object by an active method is called a "range finder". The range finder measures the distance to a target object using a camera and a projector whose relative positional relationship is known.
In this method, a light pattern whose correspondence is uniquely determined is projected from a projector onto a target object, and this is observed with a camera to obtain three-dimensional coordinates.

These rangefinders basically calculate three-dimensional coordinate values from the relationship shown in FIG. A coordinate system in which the xy plane is parallel to the image plane with the camera center as the origin,
Let's call it the camera coordinate system. The center of the projector is
Suppose that it is at (D, 0,0) T in the camera coordinate system. From the image of the light pattern projected from the projector, the three-dimensional coordinates (x, y,
z) Calculate T. At this time, the following equation is established.

(Equation 1) Here, θ is an angle between the projection direction of the light pattern and a plane including the y axis and orthogonal to the x axis.

Here, the error that occurs when the distance is measured using triangulation will be discussed. For this purpose, first, the relationship between the distance to the measurement target point and the measurement error will be analyzed using stereo vision as an example.

As shown in FIG. 16, an object point at a depth z is measured by two cameras installed at an interval D,
Parallax, that is, the positional shift amount b of the corresponding point between the two images
Is detected. If the focal length of the camera is F,
The depth z is obtained by the following equation.

(Equation 2)

[0013] The effect of a one-pixel error when parallax is obtained from an image is examined. After differentiating b in equation (2), b is eliminated using the relationship in equation (2).

(Equation 3) It becomes. Therefore, between the depth z of the measurement point and the measurement error Δz

(Equation 4) Is established.

In the laser range finder, one slit light projection device and one camera are used instead of two cameras. Assuming that the installation interval between these two devices is D, a relationship similar to Expression (4) holds.

In the present invention, based on the result of the analysis, non-uniform quantization is performed in which the quantization width of a distance image having a large distance from the sensor is set in a size proportional to the square of the distance. Generally, non-uniform quantization is performed so as to monotonously decrease with the resolution.

[0016]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS A few embodiments of the present invention will be described below with reference to the drawings. In each of the following embodiments, the quantization width is set to a value proportional to the square of the distance by using the fact that the measurement error of the range image is proportional to the square of the distance from the range image sensor. And

[First Embodiment] This is an embodiment in which an analog range image is quantized into a digital range image.
FIG. 1 shows an overall configuration diagram of a first embodiment of the present invention. In FIG. 1, reference numeral 100 denotes a range image encoding device, and 200 denotes a range image decoding device. In the distance image encoding device 100, the analog distance image measured by the distance image sensor is input, sampled into a discrete time signal by the sampling unit 110, and converted into a discrete value signal by the quantization unit 120. That is, the analog distance image is converted into digital distance image data. Generally, in the quantization, linear quantization is performed to uniformly quantize the entire amplitude axis of the image signal with a uniform width. In the quantization unit 120, the quantization width is set in each part of the object surface. In other words, non-uniform quantization for setting a quantization width in proportion to the square of the distance from the distance image sensor is performed for each part of the distance image based on the distance information. The quantized distance image is encoded by the encoding unit 130, and is stored and transmitted. In the distance image decoding device 200, after the encoded data is decoded by the decoding unit 210, the inverse quantization unit 220 performs inverse quantization based on the quantization width of each part to reproduce the distance image.

[Second Embodiment] This is an embodiment in which each pixel value of a range image is predicted from neighboring pixel values and applied to a predictive coding method for coding the residual signal. Here, as shown in FIG. 2, the distance image is obtained by converting an analog distance image measured by the distance image sensor into a discrete time signal by the sampling unit 151 and converting the analog distance image into a discrete value signal (digital distance image) by the quantization unit 152. It shall be obtained by conversion. The configuration of A / D conversion section 150 shown in FIG. 2 is basically the same as the configuration of sampling section 110 and quantization section 120 in FIG. 1, but quantum section 152 is generally used for quantization of an image signal. The difference is that a uniform quantizer is used. It is assumed that the distance images in each of the following examples are obtained by the A / D converter 150 shown in FIG. Note that a normal distance image is given by being digitized as shown in FIG.

FIG. 3 is a block diagram showing a range image coding apparatus 300 according to a second embodiment of the present invention. As a method of efficiently encoding a distance image, a method of predictively encoding a distance image can be considered. Predictive coding is one of the coding methods that focuses on the property of high correlation between adjacent pixels and improves coding efficiency.

In FIG. 3, a prediction unit 310 predicts an input distance image (digital distance image data) using a locally decoded distance image. For example, the distance image z (x,
y) is predicted by the following equation.

(Equation 5) Note that the number of local decoding distance images used for prediction can be increased. That is, the order can be increased.
The optimal order is determined by the statistical properties of the target distance image and luminance image.

The difference detection section 320 detects a residual signal between the input distance image and the prediction signal (prediction value), that is, a prediction error. Here, the residual signal is represented by the following equation.

(Equation 6) The quantization unit 330 quantizes the residual signal by, for example, the following equation.

(Equation 7) However,

(Equation 8) In the case of predictive coding, since the quantization error of the residual signal corresponds to the error of the reproduced image, δ that determines the quantization error of the residual signal is proportional to the square of the distance z as in the following equation. Set to

(Equation 9) However, the maximum value of the z _max distance image is z _max = 2 ¹⁶ −1 = 65535 for a 16-bit image. QSF (Qu
(Analysis Scale Factor)
Is a parameter for controlling the coding quality, and the smaller the QSF, the higher the quality of the stored image. When QSF is 0, lossless encoding is always performed from δ = 0, and QSF is “n”.
In this case, the quasi-reversible coding allows a maximum of “n” errors.

The variable-length encoding unit 340 entropy-encodes the quantized residual signal by variable-length encoding, and stores and transmits it.

On the other hand, inverse quantization section 350 inversely quantizes the quantized residual signal, and addition section 360 adds the inversely quantized residual signal to the prediction signal predicted by prediction section 310. . The output of the addition section 360 is fed back to the prediction section 301 and used for a local decoded distance image for the next prediction.

FIG. 4 shows a configuration diagram of a range image decoding apparatus 400 according to the second embodiment of the present invention. After the encoded data subjected to the entropy encoding is decoded by the variable length decoding unit 410, the inverse quantization unit 420 performs inverse quantization to restore the residual signal of the distance image. Inverse quantization is

(Equation 10) Is calculated by Here, the quantization width δ is calculated by Expression (9).

The signal e ′ (x, y) after inverse quantization and the residual signal e (x, y) are

[Equation 11] Holds, and the quantization error is within ± δ.

The prediction signal decoding unit 430 reproduces a distance image using the obtained residual signal and the properties of the decoded distance image near the prediction signal. The reproduced image z '(x, y) is obtained by the following equation.

(Equation 12) Here, it can be seen that the prediction signal has the same arithmetic processing in encoding and decoding, and the estimation error included in the reproduced image z ′ (x, y) is the quantization error of the residual signal itself.

[Third Embodiment] This is an embodiment applied to a method of predictive coding of a distance image by utilizing the correlation between the distance image and the luminance image. FIG. 5 shows an overall configuration diagram of the third embodiment of the present invention. In FIG. 5, reference numeral 500 denotes a range image coding device, and 600 denotes a range image coding device.

In the range image encoding device 500, the input color image is encoded by the color image encoding unit 1 and is converted into a luminance image by the luminance image extracting unit 520. If the input color image is, for example, RGB image data,
The luminance image extracting unit 520 calculates a luminance component from the R, G, and B components of each pixel to generate luminance image data. If the input color image includes a luminance plane such as YCrCb image data, the luminance image extracting unit 520 generates luminance image data by extracting only the luminance plane, that is, the Y component of each pixel. Distance image encoding section 530 receives the distance image and the luminance image, and quantizes and encodes the residual signal of the distance image in proportion to the square of the distance from the distance image sensor. The multiplexing unit 540 multiplexes the encoded data of the color image and the encoded data of the distance image, and transmits information via a communication network or a recording medium.

In the distance image coding apparatus 600, the separation unit 6
10 separates each encoded data of the color image and the distance image. The encoded data of the color image is decoded by the color image decoding unit 620 into a color image and a luminance image. The range image decoding unit 630 decodes the range image from the code data of the range image and the luminance image.

It is known that a distance image and a luminance image have a high correlation. Japanese Patent Application 2000 previously proposed by the present applicant
According to -239913, the distance image z (x, y) is
Using the signals in the vicinity and the luminance image I (x, y), the prediction can be made, for example, by the following equation.

(Equation 13) Note that the number of distance images and luminance images used for prediction can be increased. That is, the order can be increased. The optimal order is determined by the statistical properties of the target distance image and luminance image.

The residual signal is represented by the following equation.

[Equation 14] This is similar to equation (6) above. This residual signal is quantized and entropy coded. Here, the quantization is performed by equations (7) and (8), and the quantization width δ is calculated by equation (9).

The reproduced image z '(x, y) is

[Equation 15] Is calculated by Here, the prediction signal has the same arithmetic processing in encoding and decoding, and is calculated by Expression (13). Therefore, it can be seen that the estimation error included in the reproduced image z ′ (x, y) is the quantization error of the residual signal itself.

FIGS. 6 to 9 show a first configuration example of the distance image encoding unit 530 in the distance image encoding device 500 of FIG. This is an embodiment in which a distance image to be encoded is divided into blocks, and a prediction process is performed by switching a predictor and a quantization width for each block.

FIG. 6 is an overall configuration diagram of the distance image predictive encoding unit 530. The distance image block dividing unit 531, the luminance image block dividing unit 532, and the predictive encoding processing unit 5
It consists of 33. In FIG. 6, the distance image and the luminance image are divided into blocks of a fixed size by block dividing units 531 and 532, respectively. The prediction encoding processing unit 533 includes:
A prediction process is performed for each block, and the residual signal is entropy-encoded.

FIG. 7 is a diagram showing the overall configuration of the prediction code processing unit 533. The pre-processing unit 5330 and the encoded data generation unit 53
It consists of forty. FIG. 8 shows a specific configuration example of the preprocessing unit 5330, and FIG. 9 shows a specific configuration example of the encoded data generation unit 5340. The encoded data generation unit 5340 in FIG.
It is basically the same configuration.

In the predictive coding processing section 533, first, in the preprocessing section 5330, the predictive coefficient calculating section 5331 calculates a predictive coefficient using the values of the intra-block distance image and the luminance image, and a quantization step size calculating section 5332. Calculates the quantization step size (quantization width) using the value of the intra-block distance image.

Here, in the prediction coefficient calculation, a prediction coefficient for minimizing the variance of the prediction error is obtained by the least square method. Here, the right side of the equation (13) and the distance image z (x,
prediction coefficients c1, c2, c that minimize the variance of the difference y)
3 and c4 are calculated. Here, the quantization width is calculated by Expression (9). In Expression (9), the distance z is an average value of the distance from the distance image sensor using the in-block distance image.

(Equation 16) Give in.

Next, in the coded data generation unit 5340, the prediction unit 5341 generates a prediction signal by the equation (13) using the obtained prediction coefficient, and generates a difference detection unit 5342.
Generates a residual signal according to equation (14), and generates
3 quantizes the residual signal according to equations (7) and (8).
Specifically, the quantization is performed by switching the quantization width using the obtained quantization step size. Variable length coding section 534
4 entropy-codes the quantized residual signal. The residual signal (coded data) subjected to the entropy coding and the prediction coefficient and the quantization width (quantization step size) as the additional information are transmitted to the decoding side. On the other hand, the inverse quantization unit 5345 inversely quantizes the quantized residual signal, and the addition unit 5346 adds the inversely quantized residual signal to the prediction signal predicted by the prediction unit 5341 to obtain a prediction unit. Feedback to 5341.

As described above, by performing prediction processing in units of blocks, the accuracy of the prediction model is improved, so that encoding with a higher compression ratio can be performed.

FIG. 10 shows a range image decoding apparatus 60 shown in FIG.
As the distance image decoding unit 630 at 0, a configuration example corresponding to the above-described embodiment in which a prediction process is performed by switching a predictor and a quantization width for each block will be described. In FIG. 10, encoded data that has been entropy-encoded for each block is decoded by a variable-length decoding unit 631, and a quantization width (quantization step size) is calculated by an inverse quantization unit 632.
Is used to perform inverse quantization to restore the residual signal of the range image. The prediction signal decoding unit 633 decodes the block distance image using the residual signal, the block luminance image, and the prediction coefficient of the additional information.

Next, FIG. 11 shows a second configuration example of the distance image predictive encoding unit 503 in the distance image encoding device 500 of FIG. This is an embodiment in which a predicting process is performed on a range image to be encoded by switching a predictor and a quantization width for each pixel.

In FIG. 11, distance image coding section 530
Inputs a distance image and a luminance image, performs a prediction process on the distance image for each pixel, and entropy-encodes the residual signal. The overall operation is basically the same as that of FIG. 9 and will not be described. The prediction coefficient selection unit 770 selects several types of predictors prepared in advance using the properties of the decoded distance image near the prediction signal. The quantization of the residual signal is performed by Expression (8), and the quantization width δ is calculated by Expression (9). The quantization step size calculation unit 780 determines the distance z in Expression (9) using the distance from the distance image sensor estimated using the decoded distance image near the prediction signal as in Expression (17). I do.

[Equation 17]

In this model, the use of the decoded distance image near the prediction signal does not require the transmission of the additional information of the prediction coefficient and the quantization width. By performing the adaptive processing for each pixel, it is expected that the coding efficiency will be further improved as compared with the block processing.

FIG. 12 shows a range image decoding apparatus 60 shown in FIG.
As the distance image decoding unit 630 at 0, a configuration example corresponding to the above-described embodiment in which a prediction process is performed by switching a predictor and a quantization width for each pixel will be described. In FIG.
Encoded data that has been entropy-encoded for each image,
Decoding is performed in the variable length coding unit 810, and the
0, inverse quantization is performed using the predictor selected using the property of the decoded distance image near the prediction signal and the quantization width estimated using the decoded distance image near the prediction signal, and the distance image Is restored. The prediction signal decoding unit 830 decodes the distance image using the obtained residual signal and the luminance image.

[Fourth Embodiment] In a coding method for predictively coding a distance image for each block, when a signal in a block is stationary, a prediction measurement is estimated by a statistical method, and a non-stationary signal is estimated. In the case of (1), predictive coding is performed using a fixed predictor.

FIG. 13 is an explanatory diagram of this embodiment. First, it is determined whether the input distance signal in the block is stationary or non-stationary. If it is stationary, estimate the prediction coefficients using a statistical method such as the least squares method.If the non-stationarity is strong, use a fixed predictor that cannot estimate an appropriate prediction coefficient using a statistical method such as the least squares method. Used. Encoding is performed using the determined prediction coefficient. This is, for example, FIG.
In the pre-processing unit. This preprocessing further increases the coding efficiency.

Next, simulation results according to the present invention will be shown. Specifically, the coding characteristics according to the present invention were verified for a distance image having an image size of 192 × 192.
First, lossless predictive coding was performed to verify the effectiveness of the predictive coding method for range images. FIG. 14 shows the calculation results. Here, fixed prediction is a fixed predictor

(Equation 18) Is used. The block processing is a model in which prediction coefficients are estimated for each block by the least squares method.

When the unsteadiness of the signal in the block is strong, the predictor employs the equation (18). The size of the small block was 8 × 8. The entropy is 0.8882
= 0.7692 (residual signal) + 0.1058 (prediction coefficient) + 0.0132 (prediction method), which is a value including the amount of additional information such as a prediction coefficient. Here, 0.013
2 (prediction method) is an additional information amount for identifying whether a model whose prediction coefficient is estimated by the least squares method or a formula (18) is used as a predictor.

It can be seen from FIG. 14 that the entropy is reduced to 1/3 or less of the original image in any case, and that the prediction code is effective for the distance image. .

Next, FIG. 15 shows the result of quasi-reversible encoding performed on the same distance image by pixel adaptive processing. The horizontal axis indicates QSF, and the vertical axis indicates entropy.
Note that the predictor used the number (18) as in the case of the lossless encoding. FIG. 15 shows that the entropy decreases as the allowable error, that is, the QSF increases. Q
If an error is allowed up to SF = 4, lossless encoding (QSF =
0), the amount of information can be reduced to about 1/2.

[0051]

As described above, according to the present invention,
By utilizing the fact that the measurement resolution of the distance image is different at each point on the object surface, each part of the distance image is quantized with a different quantization width, so that high coding efficiency can be achieved while retaining substantial distance information. realizable.

[Brief description of the drawings]

FIG. 1 is an overall configuration diagram of a first embodiment of the present invention.

FIG. 2 is an explanatory diagram for obtaining a general distance image.

FIG. 3 is a configuration diagram of a range image encoding device according to a second embodiment of the present invention.

FIG. 4 is a configuration diagram of a range image decoding device according to a second embodiment of the present invention.

FIG. 5 is an overall configuration diagram of a third embodiment of the present invention.

FIG. 6 is an overall configuration diagram of a range image predictive encoding unit according to a third embodiment of the present invention.

FIG. 7 is a schematic configuration diagram of a prediction encoding processing unit in FIG. 6;

FIG. 8 is a specific configuration diagram of a preprocessing unit in FIG. 7;

FIG. 9 is a specific configuration diagram of an encoded data generation unit in FIG. 7;

FIG. 10 is a specific configuration diagram of a range image decoding unit corresponding to FIG. 9;

FIG. 11 is a configuration diagram of another embodiment of the range image predictive encoding unit according to the third example of the present invention.

FIG. 12 is a specific configuration diagram of a range image decoding unit corresponding to FIG. 11;

FIG. 13 is a diagram for explaining a fourth embodiment of the present invention.

FIG. 14 is a table showing specific simulation results of the present invention.

FIG. 15 is a graph showing a specific simulation result of the present invention.

FIG. 16 is a diagram illustrating the principle of measuring a distance image.

FIG. 17 is a diagram showing the principle of measuring a distance image.

[Explanation of symbols]

 REFERENCE SIGNS LIST 100 distance image encoding device 110 sampling unit 120 quantization unit 130 encoding unit 200 distance image decoding device 210 decoding unit 220 inverse quantization unit 500 distance image encoding device 510 color image encoding unit 520 luminance image extraction unit 530 Range image prediction coding unit 540 Multiplexing unit 600 Range image decoding device 610 Separation unit 620 Color image decoding unit 630 Distance image decoding unit

 ──────────────────────────────────────────────────続 き Continued on the front page F term (reference) 5C059 KK08 LB01 MA04 MA19 MA45 MD03 ME01 PP01 PP13 SS06 SS11 TA53 TB08 TB10 TC31 TC47 TD20 UA02 UA05 5C061 AA21 AA25 AB04 AB08 AB12

Claims (14)

[Claims] 1. A method of encoding a distance image having, as each pixel value, distance information to each point on an object surface measured by the distance image sensor, wherein for each part of the distance image, a distance image sensor for the part is provided. A distance image encoding method characterized in that a quantization width is set based on the resolution of the distance image, and the distance image is quantized and encoded. 2. The range image encoding method according to claim 1, wherein the quantization width of each part of the range image is set so as to decrease monotonically with respect to the height of the resolution of the range image sensor for the part. Characteristic distance image coding method. 3. The distance image encoding method according to claim 1, wherein the quantization width of each part of the distance image is set in proportion to the square of the distance between the part and the distance image sensor. Characteristic distance image coding method. 4. The distance image encoding method according to claim 1, wherein each pixel value of the distance image is predicted from adjacent pixel values, and a residual signal thereof is quantized and encoded. A range image coding method characterized by being applied. 5. The range image coding method according to claim 4, wherein the range image to be coded is subjected to a prediction process by switching a predictor and a quantization width for each pixel. Method. 6. The distance image encoding method according to claim 1, wherein each pixel value of the distance image is predicted from an adjacent pixel value and a corresponding pixel value of the luminance image, and a residual signal thereof is quantized. A range image coding method characterized by being applied to a predictive coding method to be coded. 7. The distance image encoding method according to claim 4, wherein the distance image to be encoded is divided into blocks, and a prediction process is performed by switching a predictor and a quantization width for each block. Distance image encoding method. 8. The distance image encoding method according to claim 7, wherein when the signal in the block is stationary, the prediction measurement is statistically estimated, and when the signal in the block is non-stationary, the prediction encoding is performed by a fixed predictor. A distance image encoding method. 9. A range image decoding method for decoding range image encoded data, wherein the encoded data encoded by the range image encoding method according to claim 1 is quantized into respective parts. A range image decoding method, wherein inverse quantization is performed based on a width. 10. A distance image decoding method for decoding encoded distance image data, wherein the encoded data encoded by the distance image encoding method according to claim 5 is encoded for each pixel. A distance image decoding method, comprising: performing inverse quantization using a residual signal of the obtained distance image. 11. A distance image decoding method for decoding encoded distance image data, wherein the encoded data encoded by the distance image encoding method according to claim 6 is converted to a quantization width of each part. A distance image decoding method, comprising: performing inverse quantization using a value of a residual signal and a luminance signal of an encoded distance image based on the value. 12. A distance image decoding method for decoding encoded distance image data, wherein the encoded data encoded by the distance image encoding method according to claim 7 is predicted for each block. A range image decoding method, comprising performing inverse quantization using a residual signal of the range image and the range image. 13. A distance image encoding device having a function of realizing the distance image encoding method according to claim 1. Description: 14. A range image decoding device having a function of realizing the range image decoding method according to claim 9.