GB2313245A

GB2313245A - Video signal encoder

Info

Publication number: GB2313245A
Application number: GB9710031A
Authority: GB
Inventors: Jong-Tae Lim
Original assignee: Daewoo Electronics Co Ltd
Current assignee: WiniaDaewoo Co Ltd
Priority date: 1996-05-16
Filing date: 1997-05-16
Publication date: 1997-11-19
Also published as: GB9710031D0; JPH1070723A; CN1172401A

Abstract

An Encoder encodes a video signal comprising a multiplicity of frames, wherein each frame includes object pixel block data having one or more object pixels therein and background pixel block data having no object pixel therein. The encoder comprises a transform coder 310 to transform pixel block data to provide a set of transform coefficients; a control signal generator 303 to generate a first control signal if object pixel block data is inputted and a second control signal if background pixel block data is inputted a quantization parameter controller 300 to generating a modified quantization parameter in response to either one of the first and the second control signal 5 based on the data inputted thereto; and a quantizer 320 to quantize the set of transform coefficients by using the modified quntization parameter to thereby provide a set of quantized transform coefficients.

Description

VIDEO SIGNAL ENCODER EMPLOYING AN ADAPTIVE QUANTIZATION TECHNIQUE The present invention relates to a video signal encoder; and, more particularly, to a video signal encoder capable of adaptively quantizing object and background data contained in a frame of a video signal.

In a digitally televised system such as video-telephone, teleconference or high definition television system, a video signal may be transmitted in a digital form. When the video signal including a sequence of image "frames" is expressed in a digital form, there occurs a substantial amount of digital data: for each line of an image frame is defined by a sequence of digital data elements referred to as "pixels". Since, however, the available frequency bandwidth of a conventional transmission channel is limited, in order to transmit the large amount of digital data therethrough, it is necessary to compress or reduce the volume of data through the use of various data compression techniques, especially, in the case of such low bit-rate video signal encoders as video-telephone and teleconference system.

Referring to Fig. 1, there is illustrated a conventional video signal encoder 130 for encoding a frame of a video signal. A current frame of the video signal is inputted to a subtractor 50, a motion compensation (MC) block 40 and a quantization parameter deciding block 90 on a block-by-block basis, each block having a plurality of data elements, e.g., 16x16 pixels. The encoder 130 removes or reduces spatial redundancies among pixels within block data and/or temporal redundancies between the current frame and its previous frame in order to compress the current frame to a more manageable size for transmission, wherein a transform coding technique is normally employed for reducing spatial redundancies. One of the most frequently used transform coding methods is a discrete cosine transform (DCT) method. This method is described in, e.g., Chen and Pratt, "Scene Adaptive Coder", IEEE Transactions on Communications, COM-32, No. 3, pp. 225232 (March 1984).

In an inter-mode, predicted block data for each block data of the current frame is extracted at MC block 40 by using a conventional motion estimation and compensation technique based on a previous frame stored in a memory 30 and then coupled to the subtractor 50, which generates difference block data between block data of the current frame and its corresponding predicted block data. The difference block data is then fed to a compressor 10, which includes a DCT block 11 and a quantization block 12, and is transform coded and quantized therein to generate a set of quantized transform coefficients.

The set of quantized transform coefficients is then coupled to a statistical coding block 70 which serves to generate encoded data, e.g., by using a run-length coding and a Huffman coding techniques, to be transmitted via a buffer 80 to a transmitter (not shown) for the transmission thereof.

Meanwhile, the set of quantized transform coefficients is also coupled to a decompressor 20 which includes an inverse quantization block and an inverse DCT block(not shown) and is reconstructed back to reconstructed difference block data therein.

The reconstructed difference block data is then coupled to an adder 60 wherein it is added to the predicted block data from the MC block 40 to provide a reconstructed block data of the current frame. In this way, all of the reconstructed block data for the current frame are then stored in the memory 30 to thereby form a reconstructed current frame, which ,in turn, is transmitted to the MC block 40 to be used in providing a plurality of predicted block data for processing a subsequent frame. In an intra-mode, each block data of the current frame is directly fed to the DCT block 11 and transform coded therein.

Turning now to the details of the quantization parameter deciding block 90 and the quantization block 12, in a conventional quantizer, quantization step sizes are determined based on the so-called quantization parameter (Qp) and a quantization matrix. At the quantization parameter deciding block 90, the Qp is decided for each block of, e.g., 16x16 pixels based on, e.g., the occupancy level of the buffer 80 and a variance or complexity of the current frame and transmitted to the transmitter and the quantization block 12.

The quantization block 12 having an inter-mode and an intra-mode quantization matrices in a memory (not shown) therein receives a quantization parameter Qp from the quantization parameter deciding block 90. Thereafter, base quantizer step sizes, which are the elements of a quantization matrix, are arranged so that one base quantizer step size corresponds to one of the transform coefficients of a set of transform coefficients. The quantization block 12 uses a quantization matrix and the Qp to generate the set of quantized transform coefficients as its output.

The Qp is directly related to bit rates of encoded data and the coarseness/fineness of the quantization employed in developing the encoded data. That is, a smaller Qp entails larger amount of encoded data requiring a larger number of code bits for the representation thereof, whereas a larger Qp results in smaller amount of encoded data requiring a fewer number of code bits for their representation. A larger number of code bits can represent a video signal more precisely than a fewer number of code bits. Therefore, in order to achieve a maximum picture quality under a predetermined target bit rate, it is crucial to select the Qp appropriately.

One of the coding schemes for a video signal for a low bit-rate codec (coder/decoder) system employs the so-called object oriented analysis-synthesis coding technique (see, e.g., MPEG-4 Video Verification Model Version 2.0, International Organization For Standardization ISO/IEC JTCl/SC29/WGll N 1206, March 1996), wherein a frame of the video signal is divided into a background region and/or more foregrounds or objects region which are identified by a mask image distinguishing the respect regions. In such a system, the objects can be more important than the background or, in other word, the objects may carry more crucial data than the background. Under the circumstances, therefore, it would be desirable to carry out quantization of input data adaptively depending on whether the input data belongs to the background or to the objects.

It is, therefore, a primary object of the present invention to provide a video signal encoder capable of adaptively quantizing object and background data contained in a frame of a video signal to thereby improve transmission efficiency and at the same time enhance picture quality of the video signal to be transmitted.

In accordance with the present invention, there is provided an encoder for use in a video signal encoding system, wherein the encoder encodes a video signal comprising a multiplicity of frames, each frame including object pixel block data having one or more object pixels therein and background pixel block data having no object pixel therein, to provide encoded video signal, the encoder comprising: control signal generator for generating a first control signal if object pixel block data of a frame is inputted thereto and a second control signal if background pixel block data of the frame is inputted thereto; transform coder for transform coding pixel block data, wherein the pixel block data is obtained by using either one of the object pixel block data and the background pixel block data to thereby provide a set of transform coefficients; quantization parameter controller for generating a modified quantization parameter in response to either one of the first control signal and the second control signal inputted thereto; and quantizer for quantizing the set of transform coefficients by using the modified quntization parameter to thereby provide a set of quantized transform coefficients.

The above and other objects and features of the present invention will become apparent from the following description of preferred embodiments given in conjunction with the accompanying drawings, in which: Fig. 1 illustrates a block diagram of a conventional video signal encoder; Fig. 2 shows a schematic block diagram of a video signal encoding system in accordance with the present invention; Fig. 3 depicts a detailed block diagram of an encoder shown in Fig. 2; and Fig. 4 provides a detailed block diagram of a quantization parameter control circuit depicted in Fig. 3.

Referring to Fig. 2, there is shown a schematic block diagram of a video signal encoding system 200 in accordance with the preferred embodiments of the present invention. The encoding system 200 comprises a contour coder 205, a contour decoder 210, a masking circuit 220, a background low-pass filter 230, an encoder 250 and a multiplexer (MUX) 270.

A current frame of a video signal and a contour signal of the current frame are inputted to the background low-pass filter 230 and the contour coder 205, respectively. The contour signal includes contour data representing positions of the contour pixels constituting the contour of an object in the current frame and object extraction data having information differentiating the object from a background in the current frame, wherein an object and a background pixels in the current frame are represented by, e.g., "1" and "0", respectively. At the contour coder 205, the contour data is encoded by using a conventional contour coding technique, e.g., a chain coding technique or a polygonal approximation technique, thereby providing the object extraction data and encoded contour data to the MUX 270 via a line L20. And at the same time, the encoded contour data is transmitted from the contour coder 205 to the contour decoder 210 together with the object extraction data.

At the contour decoder 210, the encoded contour data is converted into decoded contour data which, in turn is transmitted to the masking circuit 220 together with the object extraction data.

At the masking circuit 220, the contour of the object in the current frame is reconstructed as a reconstructed contour based on the decoded contour data and, further, masking information is generated based on the reconstructed contour and the object extraction data, wherein pixels residing inside the object redefined by using the reconstructed contour are represented by, e.g., l's and the remaining pixels corresponding to the redefined background are represented by, e.g., O's. The masking circuit 220 provides the masking information differentiating the redefined object pixels from the redefined background pixels contained in the current frame to the encoder 250 and the background low-pass filter 230 through a line L23.

Thereafter, based on the masking information, the background low-pass filter 230 extracts object pixel data and background pixel data from the current frame inputted thereto and then filters the background pixel data to eliminate high frequency components contained therein, thereby generating filtered background pixel data. And then the background low pass filter 230 combines the filtered background pixel data with the object pixel data, thereby providing a modified current frame having the filtered background pixel data and the object pixel data to the encoder 250 through a line L22.

It should be noted that the filtered background pixel data of the modified current frame could be encoded and transmitted with less data bits than those required in case of encoding the un-filtered background pixel data of the current frame.

The encoder 250 encodes the object pixel data and the filtered background pixel data from the background low-pass filter 230 based on the mask information from the masking circuit 220 to thereby provide encoded current frame to the MUX 270. The MUX 270 sequentially transmits the object extraction data and the encoded contour data from the contour coder 205 and the encoded current frame from the encoder 250 to a transmitter (not shown) for the transmission thereof. By virtue of transmitting the object extraction data and the encoded contour data along with the encoded current frame, the object and background can be separated at the receiving end.

According to another preferred embodiment of the present invention, the current frame may be directly inputted to the encoder 250 without low-pass filtering the corresponding background data at the background low-pass filter 230.

Referring to Fig. 3, there is depicted a detailed block diagram of the encoder 250 shown in Fig. 2. The encoder 250 includes a block formation circuit 301, a control signal generator 303, a subtractor 305, a transform circuit 310, a quantizer 320, a quantization parameter control circuit 300, a statistical coder 330, a MUX 335, a buffer 340 and a prediction circuit 325. The prediction circuit 325 has an inverse quantizer 350, an inverse transform circuit 360, an adder 370, a frame memory 380 and a motion compensation circuit 390.

The modified current frame on the line L22 and the masking information on the line L23 are inputted to the block formation circuit 301, while the quantization parameter control circuit 300 directly receives the modified current frame on the line L22. At the block formation circuit 301, the modified current frame and the masking information are divided into a plurality of pixel block data of identical size, e.g., 16x16 pixels. Each pixel block data of the modified current frame (will be referred to as "pixel block data" herein after) and its corresponding block data of the masking information (will be referred to as "masking block data" herein after) representing masking data of a pixel block data are fed on lines L33 and L30, respectively. Each pixel of masking block data is represented by, e.g., 1 or 0 depending on whether a corresponding pixel within pixel block data on the line L33 belongs to the object or the background redefined by the reconstructed contour.

Based on each masking block data on the line L30 from the block formation circuit 301, the control signal generator 303 determines whether pixel block data on the line L33 corresponds to object pixel block data or background pixel block data, wherein the object pixel block data represents pixel block data having one or more object pixels therein and the background pixel block data denotes pixel block data having no object pixel therein. According to another preferred embodiment of the present invention, the object pixel block data represents pixel block data having object pixels only therein and the background pixel block data denotes pixel block data having one or more background pixels therein. It should be also noted here that the background pixel block data is either one of filtered background pixel block data and un-filtered background pixel block data.

Specifically, the control signal generator 303 first detects zeroes included in the masking block data. Thereafter, the pixel block data on the line L33 is determined as the background pixel block data if no l's are detected in the masking block data and the object pixel block data if otherwise. The control signal generator 303 generates on a line L31 a first control signal if the pixel block data is determined as the object pixel block data and a second control signal if the pixel block data is determined as the background pixel block data.

The quantization parameter control circuit 300 generates a modified quantization parameter based on the data of the modified current frame from the background low-pass filter 230 via the line L22 in response to either one of the first control signal and the second control signal inputted thereto through the line L31, thereby providing the modified quantization parameter to the quantizer 320 and the MUX 335 via a line L32. At the motion compensation circuit 390 in the prediction circuit 325, predicted pixel block data and a motion vector for pixel block data are generated through the use of a conventional motion estimation and compensation technique, wherein most similar block data of pixel block data is searched for within a predetermined search region included in a previous frame stored in the frame memory 380 and fed to the subtractor 305 as the predicted pixel block data. The motion vector representing the displacement between the pixel block data and the most similar block data thereof is transmitted to the MUX 335 via a line L35.

At the subtractor 305, difference pixel block data or motion compensated pixel block data is obtained by subtracting the predicted pixel block data from the pixel block data. The difference pixel block data is then fed to the transform circuit 310. The transform circuit 310 transforms the difference pixel block data into a set of transform coefficients by using, e.g., a DCT technique and in turn provides the set of transform coefficients to the quantizer 320. As is well-known in the art, transformation is normally carried out with respect to a block of 8x8 pixels. The quantizer 320 quantizes the set of transform coefficients by using the modified quantization parameter fed from the quantization parameter control circuit 300 via the line L32 to thereby generate a set of quantized transform coefficients.

The set of quantized transform coefficients are then coupled to the statistical coder 330 which generates statistically coded image data by using, e.g., a run-length coding and a VLC (variable length coding) The statistically coded image data is transmitted to the MUX 335 via a line L34.

At the MUX 335, the modified quantization parameter on the line L32, the statistically coded image data on the line L34 and the motion vectors on the line L35 are multiplexed and transmitted to the buffer 340. In another preferred embodiment of the present invention, the motion vectors on the line L35 are transmitted to the statistical coder 330 first, statistically coded therein and then the statistically coded motion vectors are transmitted to the MUX 335.

The buffer 340 serves to store the multiplexed data temporarily therein and provides the stored multiplexed data to the MUX 270 shown in Fig. 2. The buffer 340 also provides state data representing the occupancy level thereof to the quantization parameter control circuit 300 via a line L36. At the quantization parameter control circuit 300, the state data is used for controlling or adjusting a modified quantization parameter to avoid overflow or underflow at the buffer 340.

In the meanwhile, the set of quantized transform coefficients are also coupled to the prediction circuit 325.

The prediction circuit 325 generates a reconstructed current frame which is to be used as the previous frame for the subsequent frame. First, in the prediction circuit 325, the inverse quantizer 350 performs inverse quantization on the set of quantized transform coefficients to thereby provide a set of inverse quantized transform coefficients to the inverse transform circuit 360. And then, the inverse transform circuit 360 performs inverse transform on the set of inverse quantized transform coefficients to thereby provide a reconstructed difference pixel block data to the adder 370.

The reconstructed difference pixel block data coupled to the adder 370 is added to the predicted pixel block data from the motion compensation circuit 390 and then reconstructed to provide reconstructed pixel block data. The reconstructed pixel block data is then stored in the frame memory 380.

In this way, all of the reconstructed pixel block data for the current frame are then stored in the frame memory 380 to thereby form a reconstructed current frame, which in turn is transmitted to the motion compensation circuit 390 to be used in providing pixel block data for processing a subsequent frame. In an intra-mode, each pixel block data of the current frame is directly fed to the transform circuit 310 and transform coded therein.

As is well-known in the art, in the case of the intramode coding, the motion compensation process for pixel block data of the modified current frame is not carried out in the motion compensation circuit 390 and the pixel block data is directly fed to the transform circuit 310 in lieu of motion compensated pixel block data; and the output of the inverse transform circuit 360 is directly coupled with the frame memory 380.

Turning now to the details of the quantization parameter control circuit 300, there is provided a detailed block diagram thereof as shown in Fig. 4 in accordance with preferred embodiments of the present invention. The quantization parameter control circuit 300 contains a quantization parameter determination circuit 401 and a quantization parameter modification circuit 402. The quantization parameter determination circuit 401 determines a quantization parameter (Qp) by using a conventional qunatization parameter determination method based on data of the modified current frame, e.g., a variance or a complexity of the modified current frame on the line L22 and the state data inputted thereto from the buffer 340 via the line L36.

And then, the Qp is transmitted to the quantization parameter modification circuit 402.

Thereafter, the quantization parameter modification circuit 402 modifies the Qp to thereby provide a modified quantization parameter (Qp') to the quantizer 320 and MUX 335 through the line L32. In detail, in response to the first control signal from the control signal generator 303 via the line L31, the quantization modification circuit 402 modifies the Qp by multiplying the Qp by K, wherein K is a predetermined number whose value is greater than 0 and smaller than 1, e.g., 0.5 to thereby provide a modified quantization parameter Qp' to the quantizer 320. And in response to the second control signal provided via the line L31, the quantization parameter modification circuit 402 modifies the Qp by multiplying the value of Qp by L, wherein L is a predetermined number whose value is greater than 1, e.g., 1.5 to thereby provide a modified quantization parameter Qp' to the quantizer 320.

Referring back to Fig. 3, the quantizer 320 receives the modified quantization parameter Qp' on the line L32 and then uses a quantization matrix (not shown) therein and Qp' to generate a set of quantized transform coefficients as its output. Therefore, in accordance with the present invention, it is possible to provide a video signal encoder for adaptively quantizing a set of transform coefficients for pixel block data in a frame of a video signal depending on whether the pixel block data belongs to the object or to the background of the frame, wherein data of the object is more finely quantized than data of the background to thereby provide enhanced picture quality under a limited target bit rate.

While the present invention has been described with respect to certain preferred embodiments only, other modifications and variations may be made without departing from the scope of the present invention as set forth in the following claims.

Claims

1. An encoder for use in a video signal encoding system, wherein the encoder encodes a video signal comprising a multiplicity of frames, each frame including object pixel block data and background pixel block data, to provide encoded video signal, the encoder comprising: control signal generating means for generating a first control signal if object pixel block data of a frame is inputted thereto and a second control signal if background pixel block data of the frame is inputted thereto; transform coding means for transform coding pixel block data, wherein the pixel block data is obtained by using either one of the object pixel block data and the background pixel block data to thereby provide a set of transform coefficients; quantization parameter control means for generating a modified quantization parameter in response to either one of the first control signal and the second control signal inputted thereto; and quantization means for quantizing the set of transform coefficients by using the modified quntization parameter to thereby provide a set of quantized transform coefficients.

2. The encoder of claim 1, wherein the object pixel block data has one or more object pixels therein and the background pixel block data has no object pixel therein.

3. The encoder of claim 1, wherein the background pixel block data is either one of filtered background pixel block data and un-filtered background pixel block data.

4. The encoder of claim 3, wherein the modified quantization parameter is obtained based on the data of the frame.

5. The encoder of claim 1, wherein each frame is divided into a plurality of equal sized blocks of M x N pixels, M and N being predetermined positive integers, respectively.

6. The encoder of claim 5, wherein the encoder further comprises: means for statistical coding the set of quantized transform coefficients to generate a statistically coded image data; and means for temporarily storing the statistically coded image data to be transmitted and for providing state data representing the occupancy level thereof.

7. The encoder of claim 6, wherein the quantization parameter control means generates the modified quantization parameter by using the state data.

8. The encoder of claim 7, wherein the quantization parameter control means includes: quantization parameter determination means for generating a quantization parameter based on the data of the frame; and quantization parameter modification means for providing the modified quantization parameter in response to either one of the first control signal and the second control signal based on the quantization parameter.

9. The encoder of claim 8, wherein the modified quantization parameter is obtained by multiplying the quantization parameter by a predetermined number K, K being a predetermined value greater than zero and smaller than 1 if the first control signal is generated from the control signal generating means and by a predetermined number L, L being a predetermined value greater than 1 if the second control signal is generated from the control signal generating means.

10. The encoder of claim 1, wherein said frames include a current frame and a subsequent frame and the encoder further comprises prediction means for generating each predicted pixel block data for each pixel block data of the subsequent frame by using said each pixel block data of the subsequent frame and a set of quantized transform coefficients corresponding to each pixel block data of the current frame.

11. The encoder of claim 10, wherein the encoder further comprises means for subtracting said each predicted pixel block data from the corresponding pixel block data of the subsequent frame to provide difference pixel block data, thereby generating sets of difference pixel block data to be transmitted to the transform coding means as sets of pixel block data of the subsequent frame.

12. The encoder of claim 10, wherein the prediction means includes: means for inverse quantizing sets of quantized transform coefficients of the pixel block data of the current frame to thereby generate sets of inverse quantized transform coefficients; means for inverse transform coding the sets of inverse quantized transform coefficients to thereby provide sets of reconstructed difference pixel block data of the current frame; means for adding said each predicted pixel block data to corresponding reconstructed difference pixel block data to provide reconstructed pixel block data, thereby generating sets of reconstructed pixel block data of the current frame; means for storing the sets of reconstructed pixel block data; and means for motion compensating by using the stored reconstructed pixel block data of the current frame from said storing means and corresponding pixel block data of the subsequent frame to thereby generate said each predicted pixel block data to be used for encoding said each pixel block data of the subsequent frame.

13. The encoder of claim 1, wherein the transform coding means is a discrete cosine transform coding means.

14. The encoder of claim 1, wherein the object pixel block data has object pixels only therein and the background pixel block data has one or more background pixels therein.

15. The encoder of claim 5, wherein M and N are 16's, respectively.

16. The encoder of claim 4, wherein the data of the frame is a variance of the frame.

17. Encoder constructed and arranged substantially as herein described with reference to or as shown in Figures 2-4 of the accompanying drawings.