KR20130078321A

KR20130078321A - Method and device for near-lossless encoding of depth image

Info

Publication number: KR20130078321A
Application number: KR1020110147192A
Authority: KR
Inventors: 최정아; 호요성
Original assignee: 광주과학기술원
Priority date: 2011-12-30
Filing date: 2011-12-30
Publication date: 2013-07-10

Abstract

Disclosed are a method and apparatus for proximity lossless depth image encoding according to the present invention.
According to an aspect of the present invention, there is provided a proximity lossless depth image encoding apparatus comprising: a difference unit which generates a difference block as a result of subtracting a prediction block from a current block; A transform unit generating discrete cosine transforms the generated difference blocks to generate discrete transformed values or transformed coefficients; And an entropy encoder that encodes all non-zero coefficients in the transformed coefficients using a unary code.

Description

Method and device for near-lossless depth image coding {Method and Device for Near-lossless Encoding of Depth Image}

The present invention relates to a method and apparatus for encoding a proximity lossless depth image. More particularly, the present invention relates to a method and apparatus for proximity lossless depth image encoding that exhibits nearly lossless performance by efficiently encoding residual data in consideration of a residual data probability distribution in a proximity lossless encoder.

Recently, as information processing technology for broadcasting and communication is rapidly developed, interest in the next generation broadcasting service is increasing. In this regard, studies on stereoscopic video and multi-view video (MVC) have been conducted to provide realistic 3D stereoscopic images that are differentiated from existing 2D images. In addition, research on free-viewpoint TV (FTV), which provides an image at any point desired by a user, is also in progress.

The depth image is an image representing 3D distance information of objects existing in the image, and represents the actual distance between the camera and the object in integer units. A virtual viewpoint other than the color image input as the depth image may be generated through the viewpoint interpolation method. Depth images are quite gentle and monotonous, unlike conventional color images.

In the 3D video coding standard, color images are encoded using a multiview video coding standard, but a standard for encoding depth images has not yet been established. Therefore, various algorithms for depth image coding have been proposed. Since the depth image is different from the color image, a coding method suitable for the characteristic of the depth image is required.

The quality of the depth image greatly affects the quality of the synthesized virtual viewpoint image. Therefore, when encoding, high quality image encoding must be performed to preserve accurate depth values. In general, a lossless encoding method is used for high quality depth image encoding.

Recently, ISO / IEC's MPEG and ITU-T's VCEG form a joint collaborative team on video coding (JCT-VC) and develop a high efficiency video coding (HEVC) standard with more than twice the compression performance of existing video compression technologies. Doing. Many domestic and international companies, research institutes and universities are participating in this standardization activity to prepare for the next generation video industry.

Currently, the 3D video coding standard group is developing a depth image coding standard based on the H.264 / AVC standard. As the HEVC standard emerges, the HEVC standard is also being considered as a platform for three-dimensional content. As the HEVC standard provides about 40% improvement in coding efficiency over the H.264 / AVC standard in color images, it is expected to perform better than the existing standard in depth images.

Therefore, it is highly likely that a depth image coding standard will be established later based on the HEVC standard. However, as with the H.264 / AVC standard, HEVC is not a standard developed for depth image coding. Therefore, in order to encode the depth image using the same, it is necessary to introduce element technology specialized in the characteristics of the depth image.

In general, when a high quality depth image is required, lossless coding using no coding loss is used. Such a lossless encoding method can decode the original depth image as it is through decoding, but has a disadvantage of low compression ratio. However, using the lossy coding method, a lot of loss occurs in the decoded depth image. Since the image quality of the depth image greatly affects the image quality of the synthesized image, loss of the depth image eventually results in deterioration of the image quality of the synthesized image.

SUMMARY OF THE INVENTION In order to solve the above problems, an object of the present invention is to provide a near lossless depth image encoding method and apparatus which can obtain a better compression ratio than a lossless depth image encoding method while obtaining a decoding result close to lossless. have.

However, the objects of the present invention are not limited to those mentioned above, and other objects not mentioned can be clearly understood by those skilled in the art from the following description.

In order to achieve the above objects, a proximity lossless depth image encoding apparatus according to an aspect of the present invention includes a difference unit for generating a difference block as a result of subtracting a prediction block from a current block; A transform unit generating discrete cosine transforms the generated difference blocks to generate discrete transformed values or transformed coefficients; And an entropy encoder that encodes all non-zero coefficients using the unary code in the transformed coefficients.

Preferably, the entropy encoder determines whether a nonzero coefficient of the current scan position exists in the transformed coefficient, and if a nonzero coefficient exists as a result of the checking, the entropy encoder encodes an absolute level using a unary code. The code of the level can be encoded.

Preferably, the entropy coding unit is implemented to use CABAC (Context-based Adaptive Binary Arithmetic Coding).

In addition, the apparatus for proximity lossless depth image encoding according to the present invention includes an inverse transform unit which deconstructs the difference block before encoding by performing inverse discrete cosine transform to use the transformed difference block for prediction of a next encoded image; And an adder configured to reconstruct the current block before encoding by adding the reconstructed difference block and the prediction block generated by intra prediction.

In addition, the apparatus for proximity lossless depth image encoding according to the present invention includes an intra prediction unit configured to predict and encode a block to be encoded using spatially adjacent pixel values in order to encode a block of an input image; And an inter predictor configured to predict the input image based on encoding information of a previous image.

In accordance with another aspect of the present invention, a method for encoding a proximity lossless depth image may include subtracting a prediction block from a current block and generating a difference block as a result of subtracting the prediction block; Discrete cosine transforming the generated difference blocks to generate discrete transformed values or transformed coefficients; And encoding a non-zero coefficient from the transformed coefficient by using a unary code.

Preferably, the step of encoding determines whether there is a non-zero coefficient of the current scanning position in the transformed coefficient, and if the non-zero coefficient exists as a result of the checking, encoding the absolute level value using a unary code. The code of the level can be encoded.

Preferably, the encoding may be performed using CABAC (Context-based Adaptive Binary Arithmetic Coding).

In addition, the method for encoding a near-lossless depth image according to the present invention includes reconstructing a difference block before encoding by performing inverse discrete cosine transform to use the transformed difference block for prediction of a next encoded image; And reconstructing the current block before encoding by adding the reconstructed difference block and the prediction block generated by intra prediction.

In addition, the method for encoding a near-lossless depth image according to the present invention includes: predictively encoding a block to be encoded using spatially adjacent pixel values to encode a block of an input image; And predicting the input image based on encoding information of a previous image.

According to the present invention, an encoding method and apparatus for a depth image capable of obtaining a lossless encoding performance with respect to a depth image and also improving a compression ratio are provided. The present invention is achieved by efficiently encoding residual data in consideration of a residual data probability distribution in a proximity lossless encoder, and can be effectively applied to depth image coding in HEVC, which is currently being research standardized.

1 is a diagram illustrating a proximity lossless coding apparatus according to an embodiment of the present invention.
2 is a diagram illustrating the existence probability of nonzero coefficients according to scan positions.
3 is a diagram illustrating a method of entropy encoding according to an embodiment of the present invention.
4 is a diagram comparing a probability distribution of absolute residual data values and an optimal probability density function of each code in a proximity lossless environment.
5 is a view showing a cutting type Golomb-Rice code.
6 is a diagram illustrating a cutting type unary code.

Hereinafter, exemplary embodiments of the present invention will be described in detail with reference to the accompanying drawings. In the drawings, the same reference numerals are used to designate the same or similar components throughout the drawings. In the following description of the present invention, a detailed description of known functions and configurations incorporated herein will be omitted when it may make the subject matter of the present invention rather unclear. In addition, the preferred embodiments of the present invention will be described below, but it is needless to say that the technical idea of the present invention is not limited thereto and can be variously modified by those skilled in the art.

In encoding a general image, a process of transforming a difference block of an input image by a method such as a discrete cosine transform and quantizing the difference block is included. In contrast, a lossless encoder has a form in which a transform and a quantization process are omitted, and a near-lossless encoder includes a transform process but a quantization process is omitted. The same is true for near lossless HEVC. In the case of near lossless HEVC, due to the irreversible nature of the transform, lossless results are not provided, but the errors that occur are negligible.

Here, the HEVC encoder is encoded using a square coding unit (CU) as a basic coding unit, and is divided into a plurality of blocks using a prediction unit (PU) as a prediction basic unit and used for prediction. . After prediction, encoding is performed using a transform unit (TU), which is a basic unit for transform and quantization. Prediction methods include an intra mode using inter-screen correlations and an inter mode based on temporal correlations between screens. In HEVC, it is proposed to consider up to 34 orientations in intra mode to improve accuracy. In addition, to improve the performance of the existing deblocking filter, the subjective picture quality improvement is considered by using an adaptive loop filter (ALF).

According to the present invention, the HEVC encoder performs the transform process without quantization and entropy encodes the transformed value in consideration of the probability distribution of the residual data. We propose a new proximity lossless HEVC coder that can achieve better compression.

1 is a diagram illustrating a proximity lossless coding apparatus according to an embodiment of the present invention.

As shown in FIG. 1, the proximity lossless coding apparatus or the proximity loss HEVC coding apparatus according to the present invention may include a difference unit 110, an intra prediction unit 120, an inter prediction unit 130, a transformer 140, and an inverse transform. The unit 150 may include an adder 160, an adder 160, a filter 170, and an entropy encoder 180.

The intra predictor 120 may predictively encode a block to be encoded using spatially adjacent pixel values in order to encode a block of an input image. In HEVC, various techniques have been proposed to improve coding efficiency. Among them, many proposals have been made regarding intra prediction. In HEVC, planar prediction and angular prediction are employed. In encoding directional blocks, each prediction is very effective.

The inter predictor 130 may predict the input image based on the encoding information of the previous image. The inter predictor 130 may include a motion predictor 132 and a motion compensator 134.

The difference unit 110 may generate a difference block by subtracting the prediction block from the current block.

The transform unit 140 may generate a discrete transformed value or a transformed coefficient by performing Discrete Cosine Transform (DCT) on the difference block generated by the difference unit 110.

The entropy encoder 180 may entropy encode encoded information such as transformed coefficients and motion vectors. The encoded information may be inserted into the bitstream and transmitted to the decoding apparatus.

Here, the entropy encoder 180 may use CABAC (Context-based Adaptive Binary Arithmetic Coding) or CAVLC (Context-based Adaptive Variable Length Coding). In particular, CABAC has a high compression ratio of about 7-14% compared to CAVLC and uses a block coding method rather than a codeword coding method. Therefore, the CABAC is preferably applied to the present invention.

The inverse transform unit 150 may reconstruct the difference block before encoding by performing inverse discrete cosine transform to use the transformed difference block for prediction of a next encoded image.

The adder 160 may reconstruct the current block before encoding by adding the reconstructed difference block and the prediction block generated by the intra predictor 120.

The filter unit 180 may be implemented with, for example, a deblocking filter, a sample adaptive offset (SAO), an adaptive loop filter (ALF), and the like, and in the proximity lossless coding environment, the quantization coefficient Since the Quantization Parameter (QP) is 0, the deblocking filter and the adaptive loop filter portion are not removed but are not actually performed.

The difference in the statistical distribution of the residual data in various encoding methods is due to the presence of transform and quantization. In the case of the proximity loss HEVC coding, since the prediction efficiency is good and no quantization is performed, the distribution of the residual data is independent of the scan position of the transformed differential block, and does not become small even when the absolute value of the residual data moves toward the high frequency side.

2 is a diagram illustrating the existence probability of nonzero coefficients according to scan positions.

As shown in FIG. 2, when various depth images are encoded in a lossy, lossless, and near lossless compression environment, the presence probability of nonzero coefficients according to scan positions may be represented in a 4x4 block.

As expected, it can be seen that the probability of occurrence of nonzero coefficients is independent of the scan position.

That is, in the proximity lossless coding method, there is no quantization process, and thus the residual data has a characteristic different from that of the conventional lossy coding method. Accordingly, in the present invention, the residual data is entropy encoded in consideration of the residual data probability distribution in the proximity lossless encoder.

3 is a diagram illustrating a method of entropy encoding according to an embodiment of the present invention.

As shown in FIG. 3, the entropy encoder according to the present invention does not encode the position of the last non-zero coefficient. This will be described in more detail as follows.

Last_significant_coeff_x and last_significant_coeff_y of CABAC syntax elements in HEVC lossy coding indicate (x, y) coordinates of the last non-zero coefficient on the scan position in the currently transformed differential block. In the case of lossy coding, since two generations of nonzero coefficients are likely to exist in the low frequency region, these two syntax elements enable efficient important map coding. However, in proximity lossless coding, encoding of two syntax elements degrades compression efficiency.

Accordingly, the present invention removes the process of encoding last_significant_coeff_x and last_significant_coeff_y, and encodes all nonzero coefficients until the end of the transformed differential block.

First, the entropy encoder may check whether a non-zero coefficient of the current scan position exists in the transformed difference block (S310).

Next, if the non-zero coefficient exists as a result of the checking, the entropy encoder may encode the absolute level value using the unary code (S320). This will be explained in detail as follows.

4 is a diagram comparing a probability distribution of absolute residual data values and an optimal probability density function of each code in a proximity lossless environment.

As shown in FIG. 4, when the proximity lossless coding is performed, the probability distribution of the absolute value coeff_abs_level_minus3 of the level and the optimal probability density function of the truncated Golomb-Rice code used in the level absolute value coding of the current HEVC are shown. . Here k is a Rice parameter of a cutting Golomb-Rice code. In near-lossless coding, the statistical properties of the absolute values of the levels approximate the optimal probability density function of the zero-order truncated Golobm-Rice code.

This zero-order truncated Golomb-Rice code is unary coded up to a symbol size of seven. The difference is that the truncated Golomb-Rice code used in HEVC has an absolute value of 8 and the Golomb code is added as a suffix. However, as shown in the figure, the absolute value of the level of the proximity lossless coding is close to the optimal probability density function of the truncated Golomb-Rice code, that is, the optimal probability density function of the unary code.

At this time, the cutting type Golomb-Rice code is shown in FIG. 5, and the unary code is shown in FIG. 6. In FIG. 5, four code tables exist according to the Rice parameter k.

Therefore, the present invention performs binarization using unary signs for absolute values of all levels. Since the present invention uses only one unary code or one unary code, there is no need to update the code table, and the complexity can be greatly improved.

Next, the entropy encoder may encode a sign of the level (S330).

Meanwhile, the above-described embodiments of the present invention can be realized in a general-purpose digital computer that can be created as a program that can be executed by a computer and operates the program using a computer-readable recording medium. The computer-readable recording medium may include a storage medium such as a magnetic storage medium (eg, a ROM, a floppy disk, a hard disk, etc.) and an optical reading medium (eg, a CD-ROM, a DVD, etc.).

The above description is merely illustrative of the technical idea of the present invention, and those skilled in the art to which the present invention pertains various modifications, changes and substitutions without departing from the essential characteristics of the present invention. This will be possible. Accordingly, the embodiments disclosed in the present invention and the accompanying drawings are not intended to limit the technical spirit of the present invention but to describe the present invention, and the scope of the technical idea of the present invention is not limited by the embodiments and the accompanying drawings. . The protection scope of the present invention should be interpreted by the following claims, and all technical ideas within the equivalent scope should be interpreted as being included in the scope of the present invention.

110: difference
120: intra prediction unit
130: inter prediction unit
140: converter
150: inverse transform unit
160: addition
170: filter unit
180: entropy encoder

Claims

A difference unit which subtracts the prediction block from the current block and generates a difference block as a result of the subtraction;
A transform unit generating discrete cosine transforms the generated difference blocks to generate discrete transformed values or transformed coefficients; And
An entropy encoder for encoding all non-zero coefficients in the transformed coefficients using a unary code;
Proximity lossless depth image encoding apparatus comprising a.

The method according to claim 1,
The entropy coding unit,
Check whether there is a non-zero coefficient of the current scanning position in the transformed coefficient,
As a result of the checking, if there is a non-zero coefficient, the absolute level is encoded using a unary code.
And a code of the level is encoded.

The method according to claim 1,
And the entropy encoder is implemented to use context-bsed adaptive binary arithmetic coding (CABAC).

The method according to claim 1,
An inverse transform unit reconstructing the differential block before encoding by performing inverse discrete cosine transform to use the transformed difference block for prediction of a next encoded image; And
An adder configured to add the reconstructed difference block and the prediction block generated by intra prediction to reconstruct the current block before encoding;
Proximity lossless depth image encoding apparatus further comprises a.

The method according to claim 1,
An intra prediction unit configured to predict and encode a block to be encoded using spatially adjacent pixel values to encode a block of an input image; And
An inter predictor configured to predict the input image based on encoding information of a previous image;
Proximity lossless depth image encoding apparatus further comprises a.

Subtracting the prediction block from the current block and generating a difference block as a result of the subtraction;
Discrete cosine transforming the generated difference blocks to generate discrete transformed values or transformed coefficients; And
An encoding unit encoding all non-zero coefficients from the transformed coefficients using a unary code;
Proximity lossless depth image encoding method comprising a.

The method of claim 6,
Wherein the encoding comprises:
Check whether there is a non-zero coefficient of the current scanning position in the transformed coefficient,
As a result of the checking, if there is a non-zero coefficient, the absolute level is encoded using a unary code.
And coding a code of the level.

The method of claim 6,
The encoding may include encoding using context-based adaptive binary arithmetic coding (CABAC).

The method of claim 6,
Reconstructing the difference block before encoding by performing inverse discrete cosine transform to use the transformed difference block for prediction of a next encoded image; And
Reconstructing the current block before encoding by adding the reconstructed difference block and the prediction block generated by intra prediction;
Proximity lossless depth image encoding method further comprises.

The method of claim 6,
Predicting encoding a block to be encoded using spatially adjacent pixel values to encode a block of an input image; And
Predicting the input image based on encoding information of a previous image;
Proximity lossless depth image encoding method further comprises.