KR100192784B1 - A video encoder - Google Patents

Abstract

The present invention relates to an image encoder, and compares a current frame of a first frame memory (44) with a previous frame of a second frame memory (46) by the image comparator (60), and then controls information and buffer control of the result. The A / D conversion control unit 62 receives the information from the unit 58 to control the sampling rate of the A / D conversion unit 40 to improve the image quality, and the image comparison unit 60 By transmitting the information about the motion-free part of the motion estimation unit 72 in advance to find the motion vector for the motion-free part during motion estimation, the computational burden can be reduced to reduce the burden on the hardware.

Description

Video encoder

1 is a block diagram showing the configuration of a conventional video encoder.

2 is a block diagram showing the configuration of an image encoder according to the present invention.

* Explanation of symbols for main parts of the drawings

40: A / D conversion unit 42: frame rearrangement unit

44: first frame memory 46: second frame memory

48: Subtractor 50: Discrete Cosine Converter

52 quantizer 54 variable length encoder

56: buffer 58: buffer control unit

60: image comparison unit 62: A / D change control unit

64: inverse quantizer 66: inverse discrete cosine transform unit

68: adder 70: third frame memory

72: motion estimation unit 74: motion compensation unit

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an image encoder, and more particularly, to an image encoder configured to control a sampling rate when a signal is converted from an analog signal to a digital signal, and to not find a motion vector when estimating a motion for a portion in which there is no motion of the image. .

In general, compression of image data is indispensable to effectively process a large amount of image information using an existing transmission band.

In particular, since a digital video signal requires a lot of memory to represent the information, the video information can be compressed, thereby efficiently storing, retrieving, and transmitting the information.

For this reason, many compression techniques for image data have been developed, and these image data compression techniques can be classified into lossless coding and lossless coding according to whether information is lost or not. ) Coding and interfrme coding using temporal redundancy present in a video.

In addition, it is divided into the first generation coding method which has little information loss and the international standard is completed, and the second generation coding method which uses the characteristics of human visual structure and image.

First-generation coding schemes include spatial coding and optimal orthogonal transformation (Karhunen-Loeve), Fourier, etc., such as pulse code modulation (PCM), differential pulse code modulation (DPCM), and delta modulation. Change coding of cosine, Harr, Hadamard, sine and the like, hybrid coding combining the above two, motion compensation coding used for video, and the like.

The second generation coding schemes include pyramid coding, anisotropic non-predictive coding, contour-texture based coding, and orientation decomposition.

Since the transmission of video has many considerations and the compression effect is more prominent than still images, the development of a technique to obtain a high compression ratio is of great interest.

HDTV (High Definition Television) broadcasting systems and Moving Picture Experts Group (MPEG) currently under development use motion compensation coding.

Motion compensation coding includes a pixel cyclic algorithm and a block matching algorithm. Although the block matching algorithm is less accurate than the pixel cyclic algorithm, it is widely used in video systems because of the simple processing and hardware implementation.

Here, the block matching algorithm divides an image into blocks of a predetermined size, obtains a motion vector for each block, and performs a discrete cosine transform (DCT) on the prediction error between the original image and the image obtained by the motion vector. It is a technique of transmitting cosine transform coefficients together.

In general, an image encoder widely used has the effect of data compression by removing temporal redundancy through motion processing and spatial redundancy through discrete cosine transform.

FIG. 1 is a block diagram showing the structure of a general video encoder. In particular, FIG. 1 shows an image encoder according to the MPEG-2 scheme.

Referring to FIG. 1, the A / D converter 10 converts an analog video signal into a digital video signal, and a sampling rate at this time is fixed.

The frame rearranging unit 12 rearranges the sequence of the images input from the camera in the order of the images for encoding, and the frame memory 14 stores the images in units of frames.

The subtractor 16 generates a difference image between frames by subtracting the motion compensation image of the original image of the current frame and the reconstructed image of the previous frame, which are input in units of frames.

The discrete cosine transforming unit (DCT) 18 outputs the discrete cosine transform coefficients by performing discrete cosine transforming of the inter-frame difference image into a block of 8x8 pixels, for example, to remove correlation between pixels.

The quantizer Q 20 quantizes the discrete cosine transform coefficients of the interframe difference image output from the discrete cosine varying unit 18 by quantizing a predetermined quantization interval.

In the variable length encoder (VLC) 22, the variable length encoding of the inter-frame difference image quantized by the quantizer 20 is performed. For example, among the signals represented by 8 bits, data with high frequency is represented by fewer bits, and the frequency is less frequent. By representing the data in many bits, the total number of bits representing the difference image is reduced.

Since the length of the data output from the variable length encoder 22 is not constant, the buffer 24 temporarily stores the data and outputs the data at a constant speed.

The buffer control unit 26 controls the quantizer 20 by determining the quantization step size according to the amount of data occupied by the buffer 24. In other words, if the data occupation amount is large, the quantization step size is increased to decrease the data amount input to the buffer 24. If the data occupation amount is small, the quantization step size is lowered to increase the data amount input to the buffer 24.

An inverse quantizer (IQ) 28 is connected to the output terminal of the quantizer 20 and restores the quantized interframe difference image to a state before being input to the quantizer 20.

The inverse discrete cosine transforming unit (IDCT) 30 is connected to an output terminal of the inverse quantizer 28, and before the inverse quantized interframe difference image is input to the discrete cosine converting unit 18 by the inverse quantizer 28. Restore to state

The adder 32 adds the inter-frame difference image and the motion compensation image reconstructed by the inverse discrete cosine transform unit 30 and stores the reconstructed image of the previous frame in the frame memory 34.

The motion estimation unit (ME) 36 generally uses a block matching algorithm, estimates a similar portion between the image of the current frame input and the image of the previous frame stored in the frame memory 34 and converts the result of the position movement into a motion vector. Output

The motion compensator (MC) 38 outputs the motion compensation image compensated by the motion vector to the subtractor 16 and the adder 32, respectively, with the motion position of the reconstructed image of the previous frame stored in the frame memory 34, respectively. do.

However, in the conventional video encoder as described above, the sampling rate of the A / D converter 10 for converting the analog video signal input from the camera into the digital video signal is not only fixed but also at a portion where there is no motion. Since the motion is also estimated, the amount of calculation of the motion estimation unit 36 increases, which causes a burden on hardware.

Accordingly, the present invention has been made to solve the above-mentioned problems, and by varying the sampling rate during A / D conversion, the image quality is improved, and the motion-free part is identified in advance so as not to find a motion vector. It is an object of the present invention to provide an image encoder that is designed to reduce the amount of computation.

According to an aspect of the present invention, there is provided an image encoder including: an A / D converter for converting an analog image signal into a digital image signal according to a variable sampling rate; A frame rearrangement unit arranged to rearrange the sequence of the image data for encoding the A / D converted image data; A first frame memory for storing the rearranged image data corresponding to a current frame on a frame basis; A second frame memory for storing image data from the first frame memory corresponding to a previous frame in units of frames; A subtractor for generating a difference image by subtracting a motion compensation image from an original image of a previous frame and a reconstructed image of a previous frame from the second frame memory; A discrete cosine transform unit for outputting a discrete cosine transform coefficient by performing discrete cosine transform to transform the difference image output from the subtractor from the spatial domain to the frequency domain; A quantizer for quantizing and outputting the discrete cosine transform coefficients of the difference image output from the discrete cosine transform unit to a predetermined quantization step size; A variable length encoder for variable length encoding and outputting the quantized difference image by the quantizer; A buffer for temporarily storing data and outputting the data at a constant speed since the length of data output from the variable length encoder is not constant; A buffer control unit configured to adjust an amount of data generation by varying a quantization section and a quantization step size for the quantization section according to the fullness of the buffer; An image comparison unit comparing a current frame from the first frame memory with a previous frame from the second frame memory to identify a portion having movement and a portion having no movement; An A / D conversion control unit adapted to vary a sampling rate in the A / D conversion unit according to the information from the image comparison unit and the information from the buffer control unit; An inverse quantizer for restoring the difference image quantized by the quantizer to a state before being input to the quantizer; An inverse discrete cosine transform unit for restoring the inverse quantized difference image by the inverse quantizer to a state before being input to the discrete cosine transform unit; An adder for adding the difference image and the motion compensation image reconstructed by the inverse discrete cosine transform unit to output the reconstructed image of the previous frame; A third frame memory for storing the reconstructed image of the previous frame output from the adder in units of frames; A similar portion is estimated between the original image of the previous frame from the second frame memory and the reconstructed image of the previous frame stored in the third frame memory, and outputs the result of the position movement as a motion vector. A motion estimator that does not search for a motion vector when information about a missing part is received; And a motion compensator for outputting, to the subtractor and the adder, a motion compensation image compensated for by a motion vector output from the motion estimator to the motion position of the reconstructed image of the previous frame stored in the third frame memory. It features.

Hereinafter, with reference to the accompanying drawings will be described in detail with respect to the present invention.

2 is a block diagram showing a configuration of an image encoder according to the present invention, wherein the image encoder includes an A / D converter 40 for converting an analog image signal into a digital image signal according to a variable sampling rate; A frame rearranging unit (42) adapted to rearrange the order of the image data for encoding the A / D converted image data; A first frame memory 44 for storing the rearranged image data corresponding to a current frame in units of frames; A second frame memory 46 for storing image data from the first frame memory 44 corresponding to a previous frame in units of frames; A subtractor 48 for generating a difference image by subtracting a motion compensation image for the original image of the previous frame and the reconstructed image of the previous frame from the second frame memory 46; A discrete cosine transforming unit (50) for outputting a discrete cosine transform coefficient by performing discrete cosine transform to convert the difference image output from the subtractor (48) into a frequency domain from a communicative region; A quantizer 52 for quantizing and outputting the discrete cosine transform coefficients of the difference image output from the discrete cosine transform unit 50 to a predetermined quantization step size; A variable length encoder (54) for variable length encoding and outputting the difference image quantized by the quantizer (52); A buffer 56 for temporarily storing data and outputting the data at a constant speed since the length of the data output from the variable length encoder 54 is not constant; A buffer control unit 58 adapted to adjust a generation amount of data by varying a quantization section and a quantization step size for the quantization section according to the fullness of the buffer 56; An image comparison unit (60) for comparing a current frame from the first frame memory (44) and a previous frame from the second frame memory (46) to identify a portion with movement and a portion without movement; An A / D conversion control unit 62 adapted to vary a sampling rate in the A / D conversion unit 40 according to the information from the image comparison unit 60 and the information from the buffer control unit 58; An inverse quantizer (64) for restoring a difference image quantized by the quantizer (52) to a state before being input to the quantizer (52); An inverse discrete cosine transform unit (66) for restoring the inverse quantized image by the inverse quantizer (52) to a state before it is input to the discrete cosine transform unit (50); An adder 68 for adding the restored source difference image and the motion compensation image by the inverse discrete cosine transform unit 66 to output the restored image of the previous frame; A third frame memory (70) for storing the reconstructed image of the previous frame output from the adder (68) in units of frames; Estimates a similar portion between the original image of the previous frame from the second frame memory 46 and the reconstructed image of the previous frame stored in the third frame memory 70, and outputs the result of the position shift as a motion vector; A motion estimator 72 that does not search for a motion vector when receiving information on a portion without motion from the image comparator 60; And a subtractor 48 and an adder to compensate for the motion position of the reconstructed image of the previous frame stored in the third frame memory 70 by the motion vector output from the motion estimator 72. And a motion compensating unit 74 for outputting the same.

Next, the operation and effects of the present invention configured as described above will be described in detail.

Referring to FIG. 2, when an analog video signal is input to the A / D converter 40, the analog video signal is converted into a digital video signal according to a variable sampling rate, and the frame is encoded to encode the order of the A / D converted image data. The rearrangement is performed in the rearrangement unit 42.

Image data corresponding to the current frame is stored in the first frame memory 44, and image data from the first frame memory 44 corresponding to the previous frame is stored in the second frame memory 46 in units of frames. It is.

The subtractor 48 subtracts the motion compensation image from the original image of the previous frame and the reconstructed image of the previous frame from the second frame memory 46 to generate a difference image.

The discrete cosine transforming unit (DCT) 50 performs discrete cosine transform to convert the difference image output from the subtractor 48 from the spatial domain to the frequency domain, and then outputs the discrete cosine transform coefficients.

After the image signal is transformed into the frequency domain by the discrete cosine transform unit 50, in general, each coefficient having a real value must be quantized to represent a limited length, that is, the quantizer 52 divides the difference image. Quantization noise occurs in the process of quantizing the cosine transform coefficients to a predetermined quantization step size, resulting in lossy coding.

The discrete cosine transform coefficient of the quantized difference image is further compressed in the variable length encoder 54 through entropy coding using a variable length code according to a statistical characteristic, which is different from the quantization process. This is no coding.

Since the length of the data output from the variable length encoder 54 is not constant, the buffer 56 temporarily stores the data and outputs the encoded bit scrim at a constant speed.

While the bandwidth of the transmission channel is fixed, the image data is finally variable-length encoded, so the amount of data generated varies with time. Therefore, there is a need for a buffer control that adjusts the amount of data generated for a given transfer rate.

In the buffer control unit 58, the amount of generation of data is adjusted by varying the step size by the quantizer 52 mainly in accordance with the fullness of the buffer 56. That is, if the number of bits generated is greater than or equal to the reference value, the amount of data filled in the buffer 56 is increased. Therefore, the quantization step size is increased to reduce the number of bits to be generated later. Adjust the state to maintain a constant value.

In this case, the quantization error occurs according to the quantization step size, which directly affects the image quality. Therefore, the buffer control technique is one of the most important factors that determine the image quality.

The image comparator 60 compares a current frame from the first frame memory 44 with a previous frame from the second frame memory 46 to identify a portion with movement and a portion without movement.

The A / D conversion controller 62 varies the sampling rate in the A / D converter 40 according to the information from the image comparator 60 and the information from the buffer controller 58. That is, the information from the buffer control unit 58 that the state of the buffer 56 is good and the information from the image comparison unit 60 that there is not much moving part of the image are transferred to the A / D conversion control unit 62. When input, the A / D converter 40 makes the sampling rate large so that the original image can be sampled densely.

In addition, the information on the sampling rate from the A / D conversion control unit 62 is transmitted to the variable length encoder 54 to recognize a change in the sampling rate during encoding.

The inverse quantizer 64 restores the difference image quantized by the quantizer 52 to a state before it is input to the quantizer 52, and the inverse discrete cosine transform unit 66 performs the inverse quantized difference image. Restores the state before input to the discrete cosine conversion unit 50.

The adder 68 adds the difference image and the motion compensation image reconstructed by the inverse discrete cosine transform unit 66 to output the reconstructed image of the previous frame, and outputs the third frame memory 70 to the adder 68. The restored image of the previous previous frame is stored in units of frames.

The motion estimation unit 72 estimates a similar portion between the original image of the previous frame from the second frame memory 46 and the reconstructed image of the previous frame stored in the third frame memory 70 to obtain a result of the position shift. When outputting as a motion vector and receiving information on a portion without motion from the image comparator 60, the motion vector is not found, so the amount of calculation in the motion estimation unit 72 is reduced.

The motion compensator 74 subtracts the motion compensation image compensated by the motion vector output from the motion estimator 72 to the motion position of the reconstructed image of the previous frame stored in the third frame memory 70. Output to the 48 and the adder 68.

As described above, the present invention can improve image quality by controlling the sampling rate to be changed when converting an analog video signal into a digital video signal. In this case, it is possible to reduce the computational burden by finding the motion vector, thereby reducing the burden on the hardware.

Claims (1)

An A / D converter 40 for converting an analog video signal into a digital video signal according to a variable sampling rate; A frame rearranging unit (42) adapted to rearrange the order of the image data for encoding the A / D converted image data; A first frame memory 44 for storing the rearranged image data corresponding to a current frame in units of frames; A second frame memory 46 for storing image data from the first frame memory 44 corresponding to a previous frame in units of frames; A subtractor 48 for generating a difference image by subtracting a motion compensation image for the original image of the previous frame and the reconstructed image of the previous frame from the second frame memory 46; A discrete cosine transforming unit for outputting a discrete cosine transform coefficient by performing discrete cosine transform to transform the difference image output from the subtractor 48 from the spatial domain to the frequency domain; A quantizer 52 for quantizing and outputting the discrete cosine transform coefficients of the difference image output from the discrete cosine transform unit 50 to a predetermined quantization step size; A variable length encoder (54) for variable length encoding and outputting the difference image quantized by the quantizer (52); A buffer 56 for temporarily storing data and outputting the data at a constant speed since the length of the data output from the variable length encoder 54 is not constant; A buffer control unit 58 adapted to adjust a generation amount of data by varying a quantization section and a quantization step size for the quantization section according to the fullness of the buffer 56; An image comparison unit (60) for comparing a current frame from the first frame memory (44) and a previous frame from the second frame memory to identify a portion with movement and a portion without movement; An A / D conversion controller 62 adapted to vary the sampling rate in the A / D converter 40 according to the information from the image comparator 60 and the information from the buffer controller 58; An inverse quantizer (64) for restoring a difference image quantized by the quantizer (52) to a state before being input to the quantizer (52); An inverse discrete cosine transform unit (66) for restoring the inverse quantized difference image by the inverse quantizer (52) to a state before being input to the discrete cosine transform unit (50); An adder 68 for adding the difference image and the motion compensation image reconstructed by the inverse discrete cosine transform unit 66 to output the reconstructed image of the previous frame; A third frame memory (70) for storing the reconstructed image of the previous frame output from the adder (68) in units of frames; Estimates a similar portion between the original image of the previous frame from the second frame memory 46 and the reconstructed image of the previous frame stored in the third frame memory 70, and outputs the result of the position shift as a motion vector; A motion estimator 72 that does not search for a motion vector when receiving information on a portion without motion from the image comparator 60; And a subtractor 48 and an adder to compensate for the motion compensation image of the reconstructed image of the previous frame stored in the third frame memory 70 by the motion vector output from the motion estimation unit 72. 68. An image encoder comprising a motion compensator 74 for output to the image.