KR100744442B1 - Improved cascaded compression method and system for digital video and images - Google Patents

Abstract

A system method for reducing quantization error in a system using a cascade compression scheme is disclosed. The expected quantization error introduced by the cascade compression scheme is determined. The expected quantization errors based on the second or higher stage test quantizers are compared. Based on this comparison, a quantizer for the second or higher stage compression is selected to reduce and / or minimize the quantization error caused by the cascade compression scheme.

Digital Video, Cascade Compression Method, Quantizer, Quantization Error, Probability Distribution Function

Description

Improved cascaded compression method and system for digital video and images

FIELD OF THE INVENTION The present invention relates generally to the field of video / image compression, and in particular, the present invention relates to systems and methods for reducing quantization errors introduced in cascaded compression of digital video and images.

Video phones, digital television, teleconferencing, and information highways are just a few of the elements of the modern digital age. Advances in digital images and video processing have helped advance the digital age. In particular, digital image compression methods have played an important role in this development. Image compression reduces the amount of data needed to represent a digital image. For example, color, gray scale, or binary images can be compressed and then decompressed to produce the original image correctly.

In general, compression is performed before storing or transmitting data. This allows a vast amount of information to be stored and / or sent quickly in an economical manner. As is apparent, image compression is generally a bidirectional process that includes compression and decompression. These processes may not be symmetrical. That is, the time and / or computational power spent on one process may vary depending upon the other type of compression algorithm used.

There are generally two types of image compression, lossy compression and lossless compression. In lossy compression, the decompressed image is similar to the original image but not exactly the same. This is because at least part of the original data is changed or disassembled. Lossy compression techniques include sample subsampling, differential pulse coding modulation (DPCM), and quantization of discrete cosine transform (DCT) coefficients. Lossless compression, on the other hand, holds all the data of the original image, which is essentially a fully reversible 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. Lossy compression methods can provide compression ratios of 100: 1 or more. Lossless compression methods generally achieve a ratio of approximately 3: 1. In general, as image loss increases as the lossy compression ratio increases, this is a compromise.

The compression ratio may be achieved by stage compression or cascaded compression of multiple stages. The image or video source is cascaded when the input source signal is multistage compressed in a serial manner. For example, in cascade compression, the source signal (image or video) is first compressed to a specific compression ratio and then to a second or higher level (s) to achieve a higher compression ratio. Indeed, higher compression may be required for efficient storage or limited bandwidth transmission.

The Joint Photographic Experts Group (JPEG) and Motion Picture Experts Group (MPEG) standards are the most commonly used compression schemes for each of images and videos. The JPEG standard is for the compression of color or gray scale images of natural real-world scenes. Although the JPEG standard includes both lossless and lossy modes, it is usually used in lossy mode to achieve greater compression ratios. Typically, images are transformed into the frequency domain using discrete cosine transform (DCT). The resulting small frequency components are removed, leaving only the larger components. These large value components are coded with differential pulse code modulation (DPCM) and Huffman coded. Since JPEG compression is an adjustable feature, variable compression ratios are possible and can also fine tune the algorithm for specific application conditions.

The MPEG standard uses DCT and Huffman coding methods along with interframe coding techniques used to provide better compression ratios. MPEG-1 and MPEG-2 are used for low resolution image sequences and high resolution sequences, respectively. MPEG-4 focuses on integrated audio-visual objects and scenes rather than frames. MPEG-7 helps to find audio-visual content.

The compression techniques achieve data compression using DCT transform followed by quantization of the DCT coefficients and variable length coding. These compression techniques are lossy compression because they quantize the DCT coefficients. As mentioned above, the lossy compression scheme is that the decompressed data is not an exact copy of the original data. For example, in the JPEG or MPEG compression scheme, each compression stage in cascade compression is in fact lossy compression. In addition, performing multiple cascade compressions introduces additional losses.

To illustrate this additional loss, reference is made to FIG. 1. Figure 1 shows two lossy compression scenarios (a) and (b). In scenario (a), source data 10 is compressed by compression system 11 in a ratio of 20: 1. In scenario (b), source data 10 is first compressed first to 10: 1 and then secondly compressed to 2: 1 by cascade compression system 12. In scenario (b), the second compression stage does not access the original source data 10 but only the compressed signal obtained as the output of the first compression stage. Nevertheless, these scenarios (a) and (b) achieve the same 20: 1 compression. However, the mean square error (MSE) introduced in scenario (b) will always be greater than or equal to scenario (a) due to cascade compression. This additional error is due in part to the selection of quantizer values at the second or higher level compression stages.

Some conventional cascade compression systems use a non-degenerative set of quantization factors for successive compression generations. This set is selected using numerical relationships such as q value = K * (3 ** n). If one of the quantizers is used in the first generation, then one of the other quantizers can be used in the subsequent generation. However, these systems do not provide any insight into how to select quantizers in subsequent generations that will minimize cascading quantization error. Thus, minimizing the MSE introduced in cascade compression, in particular the second or higher order stages of the cascaded conpression schemes, minimizes losses by proper selection of quantizers. There is a need for improved systems, methods and techniques.
WO 98/38800 discloses a cascade coding system for MPEG-2. The quantization level of the first stage quantizer is mapped to one of two possible representation levels of the second stage quantizer such that the mean squared error is reduced.

Summary of the Invention

It is an object of the present invention to solve the above limitations of the conventional cascade compression system.

It is yet another object of the present invention to provide a method for calculating the error introduced by cascade compression for a given pair of quantizers.

It is another object of the present invention to provide a method of minimizing the losses introduced by cascade compression by selecting the appropriate quantizer.

One preferred embodiment relates to reducing the quantization error introduced in JPEG and MPEG compression schemes.
According to the invention there is provided a method according to claim 1, a memory medium according to claim 9, an apparatus according to claim 13 and a system according to claim 17. Preferred embodiments are defined in the dependent claims.

One feature of the present invention is to determine an expected quantization error introduced by the second or higher stage for at least two other quantizers for the second or higher stage of the cascade compression system. And comparing the expected quantization error of the at least two quantizers. The method also includes selecting one of the quantizers according to the comparison result 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 expected quantization error.

Another aspect of the invention relates to a memory medium, an apparatus and a system for performing the method.

These and other embodiments and features of the invention are illustrated in the following detailed description.
The features and advantages of the present invention can be understood with reference to the detailed description of the preferred embodiments described by taking the following figures.

1 is a schematic diagram of a non-cascade compression system and a cascade compression system.

2 shows quantizers in a cascade compression system.

3 illustrates quantization error in a cascade compression system.

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

The additional MSE introduced by cascade compression is described with reference to the figure shown in FIG. 2 (a) shows the reconstruction points and the crystal boundaries for a uniform quantizer of step size Q ₁ . Reconstruction points are indicated by sunspots and crystal boundaries are indicated by short vertical lines. The nth reconstruction point is located at Q ⁿ ₁ (not shown) and the crystal boundaries on either side of Q ⁿ ₁ are located at D ⁿ ₁ and D ^n-1 ₁ (not shown). The crystal boundaries lie at approximately midpoint between two consecutive reconstruction points. For a uniform quantizer of step size Q ₁ , the reconstruction points lie at points that are multiples of Q ₁ . Any value of the input signal source that falls between the two decision boundaries is quantized with a reconstruction point that lies between the two decision boundaries.

Quantizer Q ₁ is used in the first stage (a) of cascade compression, and quantizer Q ₂ is used in the second stage (b) of compression. In a single compression stage, for example, a quantizer Q ₂ is used. In this embodiment, the step size of quantizer Q ₂ is shown to be larger than quantizer Q ₁ . Note that the larger the step size of the quantizer, the more loss is introduced but a higher compression ratio can be achieved. Preferably uniform quantizers such as those used in MPEG I-frames and JPEG compression schemes are used. However, other step sizes and nonuniform quantizers may also be used.

In Figure 2, x represents the value of the input source signal to be quantized. If x is in the range of [Q ⁰ ₂ , D ⁰ ₂ ), this value will be quantized to Q ⁰ ₂ by a single stage quantizer (in this case using only stage b). The "[" symbol indicates that the value is within the range and the ")" symbol indicates that the value is not within the range. In the case of the two-stage quantizer ((a) and (b)), the output of the first-stage quantizer is Q ⁰ ₁ if x is within the range of [0, D ⁰ ₁ ), and [D ⁰ ₁ , D If it is in the range of ⁰ ₂ ), the output becomes Q ¹ ₁ . The output from stage a is then passed through a second quantizer stage b that quantizes from Q ⁰ ₁ to Q ⁰ ₂ and Q ¹ ₁ to Q ¹ ₂ . Therefore, by cascade compression, x values in the range [0, D ⁰ ₁ ) are quantized to Q ⁰ ₂ and x values in the range [D ⁰ ₁ , D ⁰ ₂ ) to Q ¹ ₂ . Thus, x values in the range [D ⁰ ₁ , D ⁰ ₂ ) are quantized (incorrectly) with a larger mean squared error with the use of cascade compression compared to a single stage quantizer. Likewise, any x value in the range [D ¹ ₁ , D ¹ ₂ ) will quantize these values to Q ¹ ₂ if it is a single stage quantizer, but incorrectly to the value of Q ² ₂ by the two stage quantizers. It can be seen that it will be quantized. This additional error is a direct result of cascade compression.

In order to reduce and / or eliminate this additional error, given quantizers Q ₁ and Q ₂ , the ranges of values of x that are incorrectly quantized must first be determined. In this regard, consider a particular crystal boundary for D ⁿ ₂ to Q ₂ quantizers. In this crystal boundary, the two nearest crystal boundaries of the quantizer Q ₁ are located as shown in FIGS. 3A and 3B, one of which is D ^m ₁ greater than D ⁿ ₂ and the other D ⁿ is smaller than ₁ D ^m-1 _2.

As an example, consider the following case.

Case 1 (FIG. 3A):

A single stage Q ₂ quantizer will quantize x to Q ^{n + 1} ₂ . In the two stages of quantizers Q ₁ , Q ₂ , the Q ₁ quantizer will quantize x to Q ^m ₁ and the next Q ₂ quantizer will quantize x to Q ⁿ ₂ , thereby introducing additional quantization errors. do.

Case 2 (Figure 3b):

A single stage Q ₂ quantizer will quantize x to Q ⁿ ₂ . In the two stage quantizers Q ₁ , Q ₂ , the Q ₁ quantizer will quantize x to Q ^m ₁ , and the next Q ₂ quantizer will quantize x to Q ^{n + 1} ₂ , thereby further quantization. An error is introduced.

In order to identify all ranges of x values to be incorrectly quantized, the calculation must be repeated for all decision boundaries of Q ₂ . However, for uniform quantizers used in JPEG compression, for example, the calculation only needs to be performed for all crystal boundaries of Q ₂ in the range of 0 to LCM (Q ₁ , Q ₂ ). LCM (Q ₁ , Q ₂ ) is the least common multiple of two numbers Q ₁ and Q ₂ . The result obtained from this calculation is used to obtain ranges of x values that lie outside the range [0, LCM (Q ₁ , Q ₂ )] and are quantized incorrectly. As can be seen, the calculation is much simpler in the case of a uniform quantizer.

For each combination of pixel values in the background, the probability distribution of the black and white pixels may be different. For example, on an all white background, the probability that white pixels are encoded will be much greater than when black pixels are encoded. Given the probability distribution f (x) for the input source signal x, the expected quantization error introduced by cascade quantization using quantizers Q ₁ and Q ₂ can be calculated as follows.

The symbol ξ represents a set containing all ranges of x values that are incorrectly quantized as determined above.

The calculation of the quantization error in the above equation is used to select suitable quantizers in the continuous calculation of the second stage. For example, assuming that quantizer Q ₁ is used in the first stage and there are two possible quantizers Q ₂ , Q ' ₂ for the second stage, quantization errors for these quantizers Can be calculated. These two possible quantizers are test quantizers. The minimum quantization error value is used to determine the most suitable choice of quantizer, namely Q ₂ or Q ' ₂ .

A quantizer (Q _2), the quantization error _{E (Q 2) + E (} Q 1, Q 2) provides a bit rate r (The higher the rate, the lower the compression) having a, and a quantizer (Q _'2) are quantized If expected to provide a rate r 'with an error E (Q' ₂ ) + E (Q ₁ , Q ' ₂ ), the ratio of rate to quantization error can be used as a measure in the selection of the quantizer. Where E (Q ₂ ) and E (Q ' ₂ ) are the quantization errors inherently generated by the quantizer and are independent of the additional errors caused by cascade quantization.

Of course, quantizers other than Q ₂ or Q ′ ₂ described above may be used as the starting point. The choice of quantizers is based in part on the overall desired compression ratio. Also, some trial and error may be used in selecting early quantizers. The quantization error for several quantizers can be calculated as described above, so the most suitable quantizer is selected.

As mentioned above, one embodiment of the present invention relates to applications using JPEG and MPEG compression schemes. These compression schemes spatially partition the input source data into successive blocks of size 8 × 8 that are DCT transformed into 64 DCT coefficients. Following this, the DCT coefficients are quantized. DC coefficients are differentially coded. The 63 remaining AC coefficients are encoded by specifying the run-length of the zero coefficients and then encoding the next non-zero coefficient value.

In the case of JPEG, the entries in the quantization table determine the quantizer used for the different DCT coefficients. Different quantization tables can be used for different bands (eg, luminance and chrominance), but for a single band the quantization tables are fixed. In order to apply the quantization selection method of the present invention to recompressing already compressed JPEG data, the probability distribution f (x) for each DCT coefficient must be known. Experimentally it is known that the distribution of AC DCT coefficients follows the Laplacian distribution. Note that the parameters associated with the Laplacian distribution are different for different DCT coefficients. This parameter can be estimated, or other distributions, such as Rayliegh or Gaussian, can be obtained from the available compressed data.

It should be noted that the Laplacian distribution can be used for both luminance and chrominance channels of the DCT coded image and MPEG error terms. The MPEG compression scheme uses DCT to encode error terms as well as picture information. Error terms are obtained from the MPEG motion compensation algorithm. The error term is obtained by subtracting an image block from a block on another picture in the sequence and applying a DCT to the difference. Thus the picture can be hatched using fewer bits if there are only a few changes in the images.

In addition, in the MPEG case, the preferred embodiment focuses on I- (intra-encoded) frames in the MPEG format. I-frames are composed of only intrablocks without referring to other pictures. These frames can act as random access points in a sequence. Other MPEG frame types may also be used, such as P- (predictive encoding) and B- (bidirectional interpolation) frames.

For example, for each (I) frame in MPEG video, the quantizer value is determined by quantizer_scale and quantization table. Different quantization tables can be used for chrominance and luminance. The quantization tables are fixed for each frame, but the quantizer_scale may be changed for each macroblock. In one embodiment, quantizer_scale is selected for each MPEG frame macroblock using the quantizer selection methods described above. However, note that the quantizer_scale cannot be changed for each DCT coefficient. This value is fixed for the entire macroblock. Since the human visual system is more sensitive to quantization errors of low frequency coefficients, quantization scale is preferably selected to minimize the average quantization error of the selected low frequency coefficients.

4 illustrates a video / image processing system 20 in which the present invention may be implemented. By way of example, system 20 may be a video / image storage device such as a television, set-top box, desktop, laptop or palmtop computer, personal digital assistant (PDA), video cassette recorder (VCR), digital video recorder (DVR), TiVO device, or the like. And some or a combination thereof and other devices. 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) 22 may refer to, for example, a television receiver, a VCR or other video / image storage device. Alternatively, the source (s) 22 may comprise, for example, the Internet, wide area networks, metropolitan area networks, local area networks, terrestrial broadcast systems, cable networks, satellite networks, wireless networks, or telephone networks, and some or combinations thereof, and Other may refer to one or more network connections that receive video / images from a server or servers via a global computer communication network, such as a network type.

Input / output devices 24, processor 25 and memory 26 communicate via communication medium 27. Communication medium 27 may mean a bus, a communication network, one or more internal connections of a circuit, a circuit card or other device, and some and combinations thereof and other communication media. The input video / images from the source (s) 22 are processed in accordance with one or more software programs stored in the memory 26 and executed by the processor 25 to thereby display the display device 28, such as a television display, computer monitor, or the like. Generates output video / images to be supplied.

In a preferred embodiment, the calculation of expected quantization error due to cascade compression and the selection of suitable quantizers is implemented in computer readable code executed by system 20. The code may be stored in memory 26 or may be 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 that implement the invention.

It should be noted that the specific configuration of the system 20 as shown in FIG. 4 is merely an example. Those skilled in the art will appreciate that the present invention can be implemented using a wide variety of alternative system configurations.

Although the present invention has been described in terms of specific embodiments, it should be understood that the present invention is not intended to be limited or limited to the embodiments disclosed herein. For example, the present invention is not limited to any embodiment, compression scheme, frame type or probability distribution. On the contrary, the invention is intended to cover various modifications and variations that fall within the spirit and scope of the appended claims.

Claims (17)

In a method for a cascaded compression system (12): Determining an expected quantization error introduced by the second or higher stage for at least two other quantizers for the second or higher stage of the cascade compression system; Comparing the expected quantization error of the at least two quantizers; And Selecting one of the at least two quantizers according to a result of the comparison. The method of claim 1, And the expected quantization error is determined using a probability distribution function. The method of claim 2, And the probability distribution function is a Laplacian distribution. The method of claim 2, And said selecting step selects said one quantizer in accordance with a ratio of an expected bit rate and said expected quantization error. The method of claim 2, The probability distribution function is determined according to the input data (10) to be compressed. The method of claim 1, Wherein the at least two quantizers are determined according to the required compression ratio of the cascade compression system. The method of claim 1, The method is used in at least one of a JPEG or MPEG system. The method of claim 1, The determining step includes determining a set of ranges for values of input data to be incorrectly quantized by the cascade compression system. A memory medium containing code for the cascade compression system 12, The code is: Decision code for calculating an expected quantization error introduced by the second or higher stage for at least two other quantizers for the second or higher stage of the cascade compression system; Comparison code for comparing the expected quantization error of the at least two quantizers; And And a selection code allowing selection of one of the at least two quantizers according to a result of the comparison code. The method of claim 9, And the expected quantization error is determined using a probability distribution function. The method of claim 10, Wherein the probability distribution function is a Laplacian distribution. The method of claim 9, And the code is configured for use in at least one of a JPEG or MPEG system. In the cascade compression device 20: A memory 26 that stores executable code; And (i) determine an expected quantization error introduced by the second or higher stage for at least two other quantizers for the second or higher stage of the cascade compression system, and (ii) the at least two quantizations A processor 25 for executing the code stored in the memory 26 to compare the expected quantization error of the groups and (iii) select one of the at least two quantizers according to the result of the comparison. Cascade compression apparatus 20, including. The method of claim 13, Wherein the expected quantization error is determined using a probability distribution function. The method of claim 14, The probability distribution function is a Laplacian distribution. The method of claim 13, Wherein the device is configured for use in at least one of a JPEG or MPEG system. In the cascade compression system 12: Means (25, 26) for determining an expected quantization error introduced by the second or higher stage for at least two other quantizers for the second or higher stage of the cascade compression system; Means (25, 26) for comparing the expected quantization error of the at least two quantizers; And A cascade compression system (12) comprising means (25, 26) for selecting one of the at least two quantizers according to the result of the comparison.

Images