JP2003513563A - Improved cascade compression method and system for digital video and images - Google Patents

Abstract

SUMMARY A system and method for reducing quantization errors in a system that uses a cascaded compression scheme is disclosed. The expected quantization error caused by the cascaded compression scheme is determined. The predicted quantization errors based on two or more second or higher stage test quantizers are then compared. Based on this comparison, a quantizer for the second or higher compression stage is selected to reduce and / or minimize quantization errors caused by the cascaded compression configuration.

Description

Detailed Description of the Invention

[0001]

【Technical field】

The present invention relates generally to the field of video / image compression, and more particularly to systems and methods for reducing quantization errors that occur in digital video and image cascade compression.

[0002]

[Background technology]

Video telephony, digital television, video conferencing and information highways are just a few of the emerging digital era. Advances in digital image and video processing have helped advance into the digital age. In particular, digital image compression methods play an important role in this reform. Image compression reduces the amount of data required to represent a digital image. For example, color, grayscale or binary images can be compressed and then decompressed to produce an exact representation of the original image.

Compression is usually performed before storage or transmission of the data. This allows large amounts of information to be stored economically and / or transmitted at high speeds. As will be appreciated, image compression is a two-way process that involves compression and decompression. These processes may not be symmetrical. That is, given the type of compression algorithm used, the time taken and / or computational power for one process may differ from the other.

There are usually two types of image compression, lossy and lossless. In lossy compression, the decompressed image is similar to, but not exactly the same as, the original image. This is because at least part of the original image has been modified or discarded. Lossy compression techniques include sample subsampling, differential pulse code modulation (
DPCM) and discrete cosine transform (DCT) coefficient quantization. on the other hand,
Lossless compression preserves all of the original image data and is essentially a complete lossless encoding process. Lossless compression techniques include variable length coding (VLC) and run length coding (RLC).

The compression ratio is typically defined as the ratio between the data content to be compressed and the resulting data after compression. The lossy compression can provide compression ratios in excess of 100: 1. The lossless compression typically achieves a ratio of about 3: 1. Generally, the trade is that as the lossy compression ratio increases, so does the image degradation.

The compression ratio can be achieved by a single stage compression or a plurality of cascaded stages of compression. The image or video source will undergo cascade compression when the input source signal undergoes multiple stages of serial compression. For example, in cascade compression, the source signal (image or video) is first compressed to a particular compression ratio, and then this compressed data is subjected to a second or more (multiple) levels of compression, Get a higher compression ratio. In practice, higher compression is required for efficient storage or bandwidth limited transmission.

The Joint Photographic Experts Group (JPEG) and Motion Picture Experts Group (MPEG) standards are the most widely used compression methods for images and video, respectively. The JPEG standard is intended for compression of color or grayscale images of natural real world scenes. Although the JPEG standard includes both lossless and lossy modes, the standard is typically used in lossy mode to achieve greater compression ratios. Images are typically transformed into the frequency domain using the Discrete Cosine Transform (DCT). The resulting smaller value frequency components are rejected, leaving behind the larger value components. These large-valued components are then processed by differential pulse code modulation (DP
CM) and Huffman coding. The adjustable nature of JPEG compression allows for fine tuning of the algorithm for variable compression ratios and special application requirements.

The MPEG standard uses the DCT and Huffman coding methods with the interframe coding techniques used to produce better compression ratios. MPEG-1 and MPEG-2 are typically used for low resolution image sequences and high resolution sequences respectively. MPEG-4 focuses on unified audio-visual objects or scenes rather than frames. MPEG-7 aids in the retrieval of audio-visual content.

The compression techniques described above use a DCT transform followed by quantization of DCT coefficients and variable length coding to achieve data compression. Quantization of DCT coefficients makes these compression techniques lossy. As mentioned above, lossy compression methods are such that the decompressed data is not an exact replica of the original data. For example, JP
In the EG or MPEG compression method, each compression stage in cascade compression is lossy in nature. In addition, additional losses occur when performing multiple cascaded compressions.

To explain this additional loss, two lossy compression scenarios (A) and (B
) Is shown in FIG. In scenario (A), source data 10 is compression system 1
1 to a 20: 1 ratio. In scenario (B), source data 10
Is first subjected to a 10: 1 first stage compression by a cascaded compression system 12, followed by a 2: 1 second stage compression. In scenario (B), the second compression stage does not have access to the original source data 10, but only the compressed signal obtained at the output of the first compression stage. Nevertheless, both scenarios (A) and (B) above achieve the same compression 20: 1. However, the mean squared error (MSE) that occurs in scenario (B) will always be greater than or equal to scenario (A) due to the cascaded compression. This additional error is due in part to the choice of quantizer value in the second or higher level compression stage.

Some conventional cascade compression systems use a non-degenerative set of quantization factors for successive generations of compression. This set is selected using a numerical relationship such as q-value = K * (3 ** n). Quantizer 1
If one is used in the first generation, then any of the other quantizers can be used in subsequent generations. However, these systems do not give any insight on how to choose the quantizer that minimizes the cascaded quantization error in subsequent generations. Thus, the state of the art minimizes losses by minimizing the MSE that occurs in cascade compression, and in particular by properly selecting the quantizer in the second or higher stages of the cascade compression scheme. Such a system,
There is a need for methods and techniques.

[0012]

DISCLOSURE OF THE INVENTION

One object of the present invention is to address the limitations of conventional cascade compression systems such as those mentioned above.

Another object of the present invention is to provide a method for calculating the error caused by cascade compression for a given pair of quantizers.

Yet another object of the present invention is to provide a method for minimizing loss caused by cascade compression by selecting an appropriate quantizer.

One preferred embodiment relates to reducing quantization errors that occur within the framework of JPEG and MPEG compression methods.

One aspect of the invention is a method for a cascaded compression system, the method comprising: determining an expected quantization error caused by a second or higher stage of the cascade compression system; Comparing the expected errors of at least two quantizers for the second or higher stages. The method also includes selecting one of the quantizers based on the result of the comparison to minimize the expected quantization error for the cascade compression system.

In one preferred embodiment of the present invention, a probability distribution function is used to determine the above expected quantization error.

Another aspect of the invention relates to a memory medium and a device for implementing the method described above.

These and other embodiments and aspects of the present invention are illustrated in the detailed disclosure that follows.

The features and advantages of the present invention may be understood by referring to the detailed description of the preferred embodiments given below with reference to the drawings.

[0021]

BEST MODE FOR CARRYING OUT THE INVENTION

  An additional MSE caused by the cascade compression will be described with reference to the explanatory diagram shown in FIG.
. 2A shows Q₁Points and decision boundaries for uniform quantizers with different step sizes
It shows the world. Playback points are indicated by black circles and decision boundaries are indicated by short vertical lines.
It The nth playback point is Qⁿ ₁(Not shown), Qⁿ ₁The decision boundaries on both sides of ⁿ ₁ And D^(n-1) ₁(Not shown). Judgment boundary is approximately two consecutive
It is located halfway between the playback points. Step size Q₁On the uniform quantizer of
, The play point is Q₁Located in multiples of. Any input signal that falls between the two decision boundaries
The value of the No. source is also quantized to a playback point that falls between these two decision boundaries.

Quantizer Q ₁ is used for the first stage (A) of cascade compression and quantizer Q ₂ is used for the second stage (B) of compression. For single stage compression, for example, quantizer Q ₂ is used. In this example, the quantizer Q ₂ step size is illustrated as being larger than the quantizer Q ₁ . It should be noted that a higher quantizer step size comes at the cost of more loss, but a higher compression ratio is achieved. Preferably, uniform quantizers such as those used in MPEG I-frame and JPEG compression methods shall be used. However, other step sizes and non-uniform quantizers can be used.

[0023]   In FIG. 2, x represents the value of the input source signal to be quantized. x is
[Q⁰ _Two, D⁰ _Two), The single stage quantizer (in this case the stage (B)
Value), the value is Q⁰ _TwoWill be quantized to. Where the sign
"[" Means that the value falls within the range, symbol ")" means that the value falls within the range.
Note that it means not to enter the area. Two-stage quantizer ((A
) And (B)), x is [0, D⁰ ₁If it falls within the range of), the first stage amount
The output of the child device is Q⁰ ₁And x becomes [D⁰ ₁, D⁰ _TwoIf within the range of), Q¹ ₁ Becomes The output from stage (A) is then passed through a second quantizer (B),
The quantizer is Q⁰ ₁Q⁰ _TwoTo Q¹ ₁Q¹ _TwoTo quantize. Therefore, the cascade pressure
In case of contraction, [0, D⁰ ₁) Value x is Q⁰ _TwoIs quantized into⁰ ₁, D ⁰ _Two ) Value x is Q¹ _TwoIs quantized into. Thus, the range [D⁰ ₁, D ⁰ _Two The value of x in) is larger than that of a single-stage quantizer when using cascade compression.
(Incorrectly) quantized with a mean square error. Similarly, [D¹ ₁, D¹ _Two)of
Any value of x in the range can be Q by the two-stage quantizer.^Two _TwoIs incorrectly quantized to
However, the single-stage quantizer uses these values as Q¹ _TwoYou can see that it is quantized into. this
The additional error is a direct result of cascade compression.

[0024]   To reduce or eliminate this additional error, quantizer Q₁And Q_TwoGiven by
, The range of values of x that would be quantized incorrectly must first be determined.
No In this regard, Q_TwoD on the quantizerⁿ _TwoConsider the specific decision boundaries in
Be perceived. For this decision boundary, the quantizer Q₁The two closest decision boundaries of
That is, one is Dⁿ _TwoGreater than (D^m ₁), The other is Dⁿ _TwoLess than (D^m-1
₁), Are positioned as shown in FIGS. 3A and 3B.

[0025]   Explanatoryly, the following cases are considered.

[Equation 1]   Single stage Q_TwoThe quantizer uses x for Q^{n + 1} _TwoWill be quantized into 2-stage quantizer (Q
₁And Q_Two), Q₁The quantizer uses x for Q^m ₁Quantized into_Twoamount
Subdivider is Q for xⁿ _Two, Which results in an additional quantization error.

[Equation 2] A single stage Q ₂ quantizer will quantize x into Q ⁿ ₂ . In the two-stage quantizer (Q ₁ and Q ₂ ), the Q ₁ quantizer quantizes x into Q ^m ₁ and the subsequent Q ₂ quantizer quantizes x into Q ^{(n + 1)} _2. , Which causes additional quantization error.

This calculation must be repeated for all decision boundaries in Q ₂ in order to identify the range of all x values that will be erroneously quantized. However,
For a uniform quantizer, such as used for JPEG compression, the above calculation is 0
To LCM (Q ₁ , Q ₂ ) only need to be performed for all decision boundaries of Q ₂ . LCM (Q ₁ , Q ₂ ) is the least common multiple of the _two numbers Q ₁ and Q ₂ . The result obtained from this calculation lies outside the range [0, LCM (Q ₁ , Q ₂ )] and is used to obtain the range of values of x that are subject to false quantization. As will be appreciated, the calculation is even simpler with the uniform quantizer.

For each combination of pixel values in a situation, the probability distribution of black and white pixels may be different. For example, in an all-white situation, the probability of encoding white pixels is much greater than the probability of encoding black pixels. Given the probability distribution f (x) for the input source signal (x), the expected quantization error caused by cascaded quantization using quantizers Q ₁ and Q ₂ can be calculated as it can:

[Equation 3] The symbol ξ represents the set containing the entire range of incorrectly quantized values of x determined as described above.

The quantization error calculation in the above equation is used to select the appropriate quantizer in the second stage of the cascaded calculation. For example, assuming that quantizer Q ₁ is used in the first stage and there are two possible quantizers Q ₂ and Q ₂ 'as the second stage, the quantization error is Can be calculated. These two possible quantizers are test quantizers. The minimum quantization error value is used to determine the best choice of quantizer, Q ₂ or Q ₂ ′.

[0029]   Quantizer Q_TwoIs the quantization error E (Q_Two) + E (Q₁, Q_Two) With bit rate r
(The higher the rate, the lower the compression), and the quantizer Q _Two 'Is the quantization error E (Q_Two’) + E (Q₁, Q_TwoProvide rate r'in ')
If so, the ratio of rate to quantisation error should be chosen in the quantizer selection.
Can be used as a scale. Where E (Q_Two) And E (Q_Two’) Is quantum
Quantizer error that is originally generated by
It is not related to additional errors.

Of course, quantizers other than Q ₂ and Q ₂ ′ mentioned immediately above can also be used as starting points. The choice of quantizer is based in part on the desired overall compression ratio. Trial and error can also be used in the initial quantizer selection. The quantisation error for several quantizers can be calculated as described above, and then the most suitable quantizer is selected.

As mentioned above, one embodiment of the present invention relates to an application using JPEG and MPEG compression methods. Both of these compression methods spatially divide the input source data into contiguous blocks of 8x8 size, and subject these blocks to DCT transform, resulting in 64 DCT coefficients. This includes the DCT
Coefficient quantization follows. The DC coefficient is differentially encoded. The remaining 63 AC coefficients are encoded by identifying the run length of the zero coefficient, followed by encoding the value of the subsequent non-zero coefficient.

In the case of JPEG, the quantization table entries determine the quantizer used for different DCT coefficients. Different quantization tables can be used for different bands (eg luminance and chrominance), but these quantization tables are fixed for a single band. In order to apply the quantizer selection method of the present invention to recompress already JPEG compressed data, knowledge of the probability distribution f (x) for each DCT coefficient is required. Experimentally, it was found that the distribution of the AC DCT coefficient follows the Laplace distribution. DCs with different parameters related to Laplace distribution
Note the difference for the T coefficient. This parameter can be estimated, or another distribution such as Rayleigh or Gauss can be obtained from the available compressed data.

It should be understood that the Laplace distribution can be used for both the luminance and chrominance channels of DCT coded images as well as MPEG error terms. M
The PEG compression method uses DCT to encode error terms and image information.
The error term is obtained from the MPEG motion compensation algorithm. The error term subtracts one image block from the blocks of other images in the sequence, and the difference is DC
It is obtained by applying T. This is when the image has only a few changes,
Allows an image to be encoded with fewer bits.

Also in the case of MPEG, the preferred embodiment is I in MPEG format.
-Focus on (intra-coded: intracoded) frames. The I-frame consists only of internal blocks without reference to other images. These frames act as random access points in the sequence. Other MPs such as P- (predictively coded) frames and B- (bidirectionally interpolated) frames
EG frame types can also be used.

For example, for each (I) frame in MPEG video, a quantizer value is determined by a quantizer scale (quantizer_scale) and a quantization table. Different quantization tables can be used for chrominance and luminance.
The quantizer table is fixed for each frame, but the quantizer scale can be changed for each macroblock. In one embodiment, the quantizer selection method described above is used to select the quantizer scale for each MPEG frame macroblock. However, the quantizer scale is
Note that the T coefficient cannot be changed. This value is fixed for all macroblocks. Since the human visual system is more sensitive to the quantization error of low frequency coefficients, the quantizer scale is preferably chosen to minimize the average quantization error of the selected low frequency coefficients. To do so.

FIG. 4 shows a video / image processing system 20 in which the present invention may be implemented. Illustratively, the system 20 includes a television; a set top box; a desktop, laptop or palmtop computer; a personal computer.
A digital assistant (PDA); a video cassette recorder (VCR), a video / image storage device such as a digital video recorder (DVR), a TiVO device, etc .; and parts or combinations of these and other devices. The system 20 includes one or more video / image sources 22, one or more input / output devices 24, a processor 25, and a memory 26. Video / image source (s) 2
2 may represent, for example, a television receiver, VCR or other video / image storage device. The source (s) 22 may be, as another example, such as the Internet, a wide area network, an urban area network, a local area network, a terrestrial broadcasting system, a cable network, a satellite network, a wireless network, or a telephone network. It can represent one or more network connections that receive video / images from a server or servers via a global computer communication network, as well as parts or combinations of these and other types of networks.

The input / output device 24, the processor 25 and the memory 26 communicate via a communication medium 27. The communication medium 27 may represent, for example, a bus; a communication network; one or more internal connections of circuits, circuit cards or other devices, and parts and combinations of these and other communication media. Input video from source (s) 22 /
The image is processed according to one or more software programs stored in memory 26 and executed by processor 25 to produce an output video / image which is displayed on a television display, computer monitor, or the like. Is supplied to the device 28.

In the preferred embodiment, the calculation of the predicted quantization error due to cascade compression and the selection of the appropriate quantizer is implemented by computer readable code, such as performed by system 20. . The code can be stored in memory 26 or read / downloaded from a memory medium such as a CD-ROM or floppy disk. In other embodiments, hardware circuitry may be used in place of or in combination with software instructions to implement the invention.

It should be understood that the particular configuration of system 20 shown in FIG. 4 is for illustration only. Those skilled in the art will recognize that the present invention can be implemented with various alternative system configurations.

Although the present invention has been described in terms of particular embodiments, it should be understood that the invention is not intended to be limited or limited to the embodiments disclosed herein.
For example, the invention is not limited to any particular compression method, frame type or probability distribution. Rather, the invention is intended to cover various arrangements and modifications of the invention that are included within the spirit and scope of the appended claims.

[Brief description of drawings]

[Figure 1]   FIG. 1 is a schematic diagram of a non-cascaded compression system and a cascaded compression system.

[FIG. 2A]   FIG. 2A is an explanatory diagram of a quantizer in a cascade compression system.

FIG. 2B   FIG. 2B is also an explanatory diagram of the quantizer in the cascade compression system.

FIG. 3A   FIG. 3A is an explanatory diagram showing a quantization error in the cascade compression system.

FIG. 3B   FIG. 3B is also an explanatory diagram showing a quantization error in the cascade compression system.

FIG. 4 is a block diagram of an exemplary computer system in accordance with an aspect of the present invention.

[Explanation of symbols]

  12 ... Cascaded compression system   20. Video / image processing system   22 ... Video / image source   24 ... Input / output device   25 ... Processor   26 ... Memory   27 ... Communication medium   28 ... Display device

─────────────────────────────────────────────────── ─── Continued front page    F term (reference) 5B056 BB64                 5C059 KK01 MA00 MA05 MA14 MA23                       MC11 PP05 PP06 PP07 SS02                       SS06 SS07 TA46 TB07 TC08                       TD02 TD06 TD16 UA02                 5C078 AA04 BA57 CA22 DA01                 5J064 AA02 BB03 BC16 BC28 BD01

Claims (17)

[Claims] 1. A method for a cascaded compression system, the method comprising: determining a predicted quantization error caused by a second or higher stage of the cascaded compression system; Comparing the predicted quantization error of at least two quantizers for higher stages, and selecting one of the at least two quantizers based on the result of the comparison. A method for a cascaded compression system characterized by. 2. The method of claim 1, wherein the predicted quantization error is determined using a probability distribution function. 3. The method according to claim 2, wherein the probability distribution function is a Laplace distribution. 4. The method of claim 2, wherein the selecting step selects the one quantizer based on a ratio of a predicted bit rate and the predicted quantization error. How to characterize. 5. The method according to claim 2, wherein the probability distribution function is determined based on the input data to be compressed. 6. The method of claim 1, wherein the at least two quantizers are determined based on a desired compression ratio of the cascaded compression systems. 7. The method of claim 1, wherein the cascaded compression system comprises at least one of a JPEG system or an MPEG system. 8. The method of claim 1, wherein the determining step comprises
A method comprising: determining a series of ranges of values of input data that will be erroneously quantized by the cascaded compression system. 9. In a memory medium containing a code for a cascaded compression system, the code calculates a predicted quantization error caused by a second or higher stage of the cascaded compression system. A decision code, a comparison code comparing the predicted quantization errors of at least two quantizers for the second or higher stages, and a comparison code for the at least two quantizers based on a result of the comparison code. A memory medium, comprising: a selection code enabling one selection. 10. The memory medium of claim 9, wherein the predicted quantization error is determined using a probability distribution function. 11. The memory medium according to claim 9, wherein the probability distribution function has a Laplace distribution. 12. The memory medium according to claim 9, wherein the cascaded compression system has at least one of a JPEG system and an MPEG system. 13. A cascaded compression device, a memory storing executable code, and executing the code stored in the memory to (i) a second or a second cascaded compression system. Determining the predicted quantization error caused by the higher stage, (ii) comparing the predicted quantization error of at least two quantizers for the second or higher stage, and (iii) Based on the result of the comparison, the at least 2
A cascaded compression device, comprising: a processor enabling selection of one of the four quantizers; 14. The apparatus according to claim 13, wherein the predicted quantization error is determined using a probability distribution function. 15. The apparatus according to claim 14, wherein the probability distribution function is a Laplace distribution. 16. The apparatus according to claim 13, wherein the cascaded compression system includes at least one of a JPEG system and an MPEG system. 17. In a cascaded compression system, means for determining an expected quantization error caused by the cascaded compression system, and a second or higher stage of the cascaded compression system. Means for comparing the predicted quantization error based on the possible quantizers, and selecting one of the possible quantizers to reduce the predicted quantization error A cascading compression system comprising: