KR100311573B1 - Wavelet-based software encoder and decoder - Google Patents

Abstract

PURPOSE: Wavelet-based software encoder and decoder are provided to prevent the block effect of a decoded image and to obtain a compressed image proper to multi-channel transfer by using the characteristic of a wavelet. CONSTITUTION: A wavelet-based software encoder comprises a wavelet transforming unit(11) for splitting an input image by each band employing a wavelet filter, an average energy calculating unit(12) for calculating the average energy of the images of respective bands, a decision control unit(13) for deciding the priority order of bands to transfer depending on the average energy, a scalar quantization unit(14) for scalar-quantizing the image of the highest band, and a run-length coding unit(16) for run-length coding the images on the other bands and transferring the images. A wavelet-based software decoder comprises a receiving unit for receiving the compressed image of respective bands from the encoder, a decoding unit for decoding the compressed images, and a wavelet inverse converting unit for regenerating the decoded images as an original image.

Description

Wavelet-based Software Encoder and Decoder

The present invention relates to a wavelet service in a service requiring multiple transport channels according to a network or a quality of service (QOS) that supports priority for stream transmission, such as an asynchronous transfer mode (ATM). The present invention relates to a wavelet-based software encoder and decoder suitable for hierarchical transmission by compressing and reconstructing an image using a feature, and capable of operating in real time.

Generally, software-based encoders widely used on the Internet include nv (Network Video), CUSee-Me, and IVS.

The nv is a software encoder developed by LBL, has a compression ratio of about 18: 1 to 20: 1, divides an image into 8 × 8 block units, and marks blocks that have changed above a threshold in each block of the current frame and the previous frame. After this, two-dimensional Haar transform or DCT is performed, variable length coding is performed, and at this time, an algorithm for periodically transmitting unchanged blocks is used to improve image quality.

The CUSee-Me converts the video into 4 bits of 16-gray, divides it into 8 × 8 blocks, and compresses and transmits only the most changed part of the previous frame and the current frame, and the periodic transmission is the same as the nv. do.

In IVS, the H.261 standard is implemented by software, and it is divided into two parts: block analysis and frequency analysis.

That is, the first process divides the image into 8 × 8 blocks, and if each block does not change much compared to the corresponding block of the previous image, this block is not transmitted. If there is a change, it transmits the changed amount. It can also be a motion vector.

The second process performs DCT, quantizes it, and then performs variable length coding to transmit it.

The advantage of these encoders is that they are compatible with other H.320-compliant video encoders, and the compression ratio is quite high.

In addition, the recently developed UCB / LBL vic takes advantage of nv and IVS by designing the Intra-H.261 coder that combines the above nv and IVS.

However, in the above-described conventional software encoder, a block effect occurs at the receiving side due to a common characteristic of an algorithm that divides an image into 8 × 8 blocks, processes each block, and then compresses and transmits the image. There was this.

In addition, since the conventional software encoder is a single layer encoder, there is a problem in that it is not possible to provide a compressed image necessary for transmission through multiple channels.

SUMMARY OF THE INVENTION The present invention has been made to solve the above problems, and an object thereof is to provide a wavelet-based software encoder suitable for hierarchical transmission and to operate in real time by compressing and reconstructing an image using the characteristics of the wavelet, and To provide a decoder.

In order to achieve the above object, the wavelet-based software encoder and decoder of the present invention divides an input image into bands by using a wavelet filter, and then obtains an average energy of the image divided into the bands. The video is transmitted by performing variable length coding after scalar quantization, and the remaining video is transmitted by performing run-length coding. The decoder decodes the compressed video transmitted by the encoder and then inversely transforms the original wavelet. Characterized in that to restore to the image.

1 is a block diagram of a wavelet-based software encoder according to the present invention;

2 is a block diagram of a wavelet-based software decoder according to the present invention;

3 is a diagram illustrating a transmission form of a layered image of a wavelet-based software encoder according to an embodiment of the present invention.

<Description of the symbols for the main parts of the drawings>

11 wavelet converter 12 band average energy calculation unit

13: Determination Control Unit 14: Scalar Quantization Unit

15 variable length coding unit 16 run-length coding unit

21: receiver 22: decoder

23: wavelet inverse transform unit

Hereinafter, the configuration and operation of a wavelet-based software encoder and decoder according to the present invention will be described in detail with reference to the accompanying drawings.

1 is a block diagram of a wavelet-based software encoder according to the present invention, Figure 2 is a block diagram of a wavelet-based software decoder according to the present invention.

As illustrated in FIG. 1, a wavelet-based software encoder divides an input image into bands by using a wavelet filter, and a wavelet transform unit 11 and band-by-band splits by the wavelet transform unit 11. A band-specific average energy calculator 12 for calculating an average energy of an image, a determination controller 13 for determining a priority of a band to be transmitted according to energy obtained by the band-specific average energy calculator 12, and the determination Scalar quantization unit 14 for scalar quantizing an image of a band determined by the controller 13 to have the largest energy, and variable length coding for variable length coding the quantized image in the scalar quantization unit 14 and transmitting the same. Run-Length for run-length coding the image of the band determined by the variable length coding 15 and the determination controller 13 to have the second and third largest energy. It is composed of a coding unit (Run Length Coding) 16.

The wavelet-based software decoder illustrated in FIG. 2 includes a receiver 21 for receiving a compressed image compressed and transmitted by an encoder for each band and a decoding unit for decoding the compressed image received by the receiver 21 ( 22) and a wavelet inverse transform unit 23 for combining the band-decoded images decoded by the decoder 22 to restore the original image.

The wavelet transform in the present invention can represent any function f (x) as a linear combination of wavelet basis functions having locality simultaneously in time-frequency space.

The operation of the wavelet-based software encoder and decoder configured as described above is as follows.

First, the wavelet transform unit 11 of the encoder performs 2 to 3 levels of wavelet transform on the input image, and then removes coefficients below the threshold.

At this time, the image is divided into bands using a wavelet filter, and a bi-orthogonal wavelet filter is used.

In other words, since the DC wavelet filter has a nonlinear phase characteristic, phase distortion occurs, and thus the image wavelet filter is poor in image processing, a bi-orthogonal wavelet filter is used.

The bi-orthogonal wavelet filter has a characteristic of reducing the amount of computation since the operation can be performed by a sum and a difference with a shift excluding the floating-point arithmetic, and is expressed as follows.

H_O (Z) = -0.0625 Z **-2 + 0.0625 Z **-1 + 0.5 + 0.5 Z + 0.0625 Z ** 2-0.0625 Z ** 3

Subsequently, the image divided for each band by the wavelet converter 11 is input to the average energy calculator 12 for each band, and accordingly, the average energy calculator 12 for each band averages each subband of the image. Calculate your energy.

Then, the determination controller 13 arranges the energy obtained by the average energy calculator 12 for each band in ascending order and takes only three bands in the largest order.

Accordingly, since the portion determined to have the greatest energy from the determination controller 13 includes the most important component in the current image, scalar quantization is performed through the scalar quantization unit 14 having a low loss rate, and then variable length coding. The variable length coding is performed through the unit 15 and then transmitted.

In addition, the image of the remaining band is transmitted from the determination controller 13 through the run-length coding unit 16 to perform zero crossing run-length coding.

At this time, in the compression of the actual image, a pyramid wavelet transform having a fast execution speed is used.

Meanwhile, the wavelet transform unit 11 of the encoder divides the input image into bands as shown in FIG. 3A, and each band is represented by a constant between 0 and 7. FIG.

That is, after the image is divided into two levels, 0 is transmitted using LL (Low Low) 2.

In addition, in order to use the encoder according to the present invention in a video conference system, since the video compression scheme has a hybrid form, information on the compression scheme is required to restore the compressed image at the reception side.

That is, as shown in (b) of FIG. 3, an image of a location (1), a compression mode (2), a width (3), and a height (4) of an image Compressed image is transmitted along with the compressed information.

Subsequently, the decoder which reconstructs the compressed image encoded and transmitted by the encoder to the original image receives the band-specific compressed image through the receiver 21 and outputs the compressed image to the decoder 22.

Then, the decoding unit 22 decodes the compressed image received by the receiver 21, and then combines the original band-specific image decoded by the decoder 22 in the wavelet inverse transform unit 23. Restore to.

As described above, the present invention has an effect of preventing the block effect of the reconstructed image by using the characteristics of the wavelet and obtaining a compressed image suitable for multi-channel transmission.

Claims (2)

A wavelet converter 11 for dividing the input image into bands by using a wavelet filter, and an average energy calculator for each band, which calculates an average energy of the band-specific image divided by the wavelet converter 11 ( 12) and a decision controller 13 for determining the priority of a band to be transmitted according to the energy obtained by the average energy calculator 12 for each band, and an image of a band determined by the decision controller 13 to have the greatest energy. The scalar quantization unit 14 for scalar quantization, the variable length coding unit 15 for transmitting variable length coding of the quantized image in the scalar quantization unit 14, and the second and third in the decision controller 13 An encoder comprising a run-length coding unit 16 for performing run-length coding and transmitting an image of a band determined to have high energy; A receiver 21 for receiving a compressed image compressed and transmitted by the encoder for each band, a decoder 22 for decoding the compressed image received by the receiver 21, and a band decoded by the decoder 22. A wavelet-based software encoder and decoder, comprising: a decoder comprising a wavelet inverse transform unit (23) for combining a star image to restore an original image. The wavelet-based software encoder and decoder of claim 1, wherein the wavelet filter of the wavelet transform unit is a bi-orthogonal wavelet filter.

Images