KR102020953B1 - Image Reencoding Method based on Decoding Data of Image of Camera and System thereof - Google Patents

Abstract

The video recoding system based on the decoding information of the camera image of the present invention receives a compressed stream transmitted from a camera, outputs a reconstructed image, and transmits the reconstructed image to the image encoding unit. A re-encoding determiner for receiving the decoded information transmitted from the decoder, determining a re-encoding method with reference to the decoded information, and transmitting the determined re-encoding method to the image encoder, and receiving a reconstructed image from the image decoder, Re-encoding the reconstructed image by referring to the received re-encoding method. Re-encoding the image by referring to the bandwidth (bit rate) and image quality (distortion) information received from the distortion measuring unit, and rate the re-encoded image and the compressed stream. An image encoder for transmitting the distortion measurement unit, and an image encoder Rate for transmitting the measured rotor bandwidth (bit rate) of the re-encoded picture quality and compression of the video stream and measures the received quality and the bandwidth (bit rate), image coding portion-to-part wherein the distortion measure.

Description

Image Reencoding Method Based on Decoding Information of Camera Image and Image Recoding System Using the Same [Image Reencoding Method based on Decoding Data of Image of Camera and System]

Due to the development of cameras, display devices, and storage devices, the resolution of video contents on the market is continuously increasing. In the past, content of SD (720x480) resolution of about 300,000 pixels was mainly used, but content of FullHD (1920x2160) resolution of more than 2 million pixels and UHD (3840x2160) and 4K (4096x2160) level of over 8 million pixels Resolution content is actively being distributed. In addition, with the development of mobile terminals, the demand for high-definition video services anytime, anywhere is increasing. Therefore, it is necessary to reduce the burden of bandwidth on the network while serving the above high quality video.

The related art related to the present invention is disclosed in Korean Patent Publication No. 10-2016-0051356 (published on May 11, 2016). 1 is a block diagram of the conventional image re-encoding apparatus. In FIG. 1, the decoder 1300 may include an entropy decoder 1310, an inverse quantizer 1320, and an inverse transformer 1330. The entropy decoder 1310 parses the encoded image file 1305 to obtain an encoded image to be decoded and encoding information necessary for decoding. The entropy decoder 1310 may obtain a first quantization matrix for dequantizing the encoded image by entropy decoding a header of the encoded image file 1305. Also, the entropy decoder 1310 may entropy decode the encoded image of the encoded image file 1305 to obtain a quantized transform coefficient generated in units of blocks. The inverse quantizer 1320 inversely quantizes the quantized transform coefficients generated in units of blocks by using the first quantization parameter included in the first quantization matrix to restore the transform coefficients. The inverse transformer 1330 inversely transforms the transform coefficients reconstructed by the inverse quantizer 1320 to obtain a reconstructed image 340. The inverse transform unit 1330 may inversely transform the transform coefficients by an inverse discrete cosine transform. The reconstructed image file obtained by the decoder 1300 is input to the recoder 1350. The recoding unit 1350 may include a transform unit 1360, a quantization unit 1370, and an entropy encoder 1380. The converter 1360 may output the transform coefficients generated by converting the reconstructed image 1340 in units of blocks. For example, the transform unit 1360 may output a transform coefficient by performing discrete cosine transform on the reconstructed image in units of 8 × 8 pixels. The quantization unit 1370 may quantize the transform coefficients output from the transform unit 1360 using the second quantization matrix and output the quantized transform coefficients to the entropy encoder 1380. The entropy encoder 1380 may generate the recoded image file 1395 by performing entropy encoding on the quantized transform coefficients. When entropy encoding is performed, a small number of bits may be allocated to a symbol having a high frequency of occurrence and a large number of bits may be allocated to a symbol having a low frequency of occurrence. Therefore, in this case, the length of the bitstream including the bit strings representing the symbols may be shorter than in the case of randomly assigning the bit strings to the symbols. Therefore, compression performance of image encoding may be increased through entropy encoding. The entropy encoder 1380 according to an embodiment may use a variable length coding (VLC) encoding method for entropy encoding.

2 is another embodiment of a conventional video re-coding apparatus. In FIG. 2, the conventional video recoding apparatus is composed of an image decoding apparatus and an image encoding apparatus, and the image decoding apparatus is a module that decodes a syntax from compression (binary) by recovering an image by receiving a compressed stream. The inverse quantization / inverse transform unit, which is a module for restoring the residual signal by inversely transforming an effective coefficient (an important coefficient of the residual signal) among the syntaxes decoded by the decoder and the entropy decoder, uses a spatial correlation of pixels A module for reconstructing. A spatial predictor for predicting a current pixel value from neighboring pixel values of a currently decoded pixel. A module for reconstructing a temporal overlapping signal using temporal correlation of pixels. Motion compensation to predict the value of the decoded pixel And refer to the decoded information of the restored pixel due to the pixel value which a module for filtering (change) that would be composed of a filtered image that the image quality is improved. In addition, the video encoding apparatus receives an image and outputs a compressed stream. The video encoding apparatus removes a spatial redundancy signal by using a spatial correlation of pixels, and predicts a current pixel value from adjacent pixel values of a pixel to be decoded. A module that removes a temporal redundant signal using a temporal correlation of pixels, and includes a motion estimator, a spatial predictor, or a motion estimator that predicts a value of a pixel to be currently decoded from a frame temporally adjacent to a currently decoded frame. A transform / quantizer which is a module for transforming the residual signal from which the redundant signal is removed into a frequency domain and then quantizes and compresses it, an entropy encoder which is a module that encodes the syntax determined by the effective coefficient and the spatial prediction unit and the motion estimation unit after the transform quantization unit, The pixel value is referred to by referring to encoding information of the encoding pixel. Sintering (change) see a module for Therefore storing the temporally adjacent frames to be referenced by the image filtering unit and the motion estimation quality is improved, which will be of a video buffer. In the case of H.264, the entropy decoding unit has an entropy decoding method such as CAVLC and CABAC. In addition, in the case of H.264, the entropy encoder includes an encoding method such as CAVLC and CABAC. In the conventional video re-encoding apparatus configured as described above, the reconstructed image, which is the decoding output, becomes the input image of the encoding apparatus, and the compressed stream, which is the output of the decoding apparatus, and the compressed stream, which is the output of the encoding apparatus, are encoded by different encoding schemes. There is no connection between the two.

The conventional video recoding method configured as described above has a problem of increasing the transmission bandwidth of the system by increasing the bandwidth of the image when the resolution of the image content is increased. In addition, the conventional re-encoding method as described above has a problem that it is difficult to provide high-definition image content to a plurality of users anytime, anywhere the image transmitted from a plurality of cameras. Accordingly, an object of the present invention is to provide a method for reducing video bandwidth while transmitting high quality video contents to a plurality of users in a surveillance system to which a plurality of cameras are connected.

The video recoding system based on the decoding information of the camera image of the present invention having the above object receives the compressed stream transmitted from the camera, outputs a reconstructed image, and transmits the reconstructed image to the image encoding unit, and outputs the decoding information to the recoding determination unit. A decoding unit, a recoding determination unit receiving the decoding information transmitted from the image decoding unit, determining a recoding method with reference to the decoding information, and transmitting the determined recoding method to the image encoding unit, and receiving a reconstructed image from the image decoding unit. And re-encoding the reconstructed image by referring to the re-encoding method received from the re-encoding determination unit, re-encoding the image by referring to the bandwidth (bit rate) and image quality (distortion) information received from the rate-distortion measurement unit. To transmit the compression and compression stream to the rate-distortion measurement unit. And a rate-distortion measurement unit for measuring the image quality and the bandwidth (bit rate) of the compressed stream received from the image encoder and transmitting the measured image quality and bandwidth (bit rate) to the image encoder. will be.

The video recoding method based on the decoding information of the camera image and the video recoding system using the same according to the present invention configured as described above can reduce the time required for video recoding, thereby enabling high speed execution, and also target bandwidth (bit rate) of the recoded video. ) Will provide the best picture quality. In addition, another effect of the present invention is to reduce the bandwidth burden of the network in a surveillance system to which a plurality of cameras are connected to provide a smooth video service environment.

1 is a block diagram of a conventional image re-encoding apparatus;
2 is a configuration diagram of another conventional image re-coding apparatus;
3 is a block diagram of a video recoding system based on decoding information of a camera image according to the present invention;
4 is a control flowchart of a decoding information-based image re-coding method of a camera image of the present invention;
5 is a control flowchart of a re-coding method of a low quality area of a reconstructed image applied to the present invention.

A video recoding method based on the decoding information of the camera image of the present invention having the above object and a video recoding system using the same will be described below with reference to FIGS. 3 to 5.

3 is a block diagram of a decoding information based video re-coding system of a camera image according to the present invention. In FIG. 3, the decoding information-based image recoding system of the present invention includes receiving a compressed stream transmitted from a camera, outputting a reconstructed image, and transmitting the reconstructed image to the image encoding unit, and outputting decoding information to the recoding determination unit. A re-encoding determination unit 200 which receives the decoded information transmitted from the image decoding unit, determines a re-encoding method with reference to the decoded information, and transmits the determined re-encoding method to the image encoding unit, and from the image decoding unit. Re-encoding the reconstructed image by referring to the re-encoding method received from the re-encoding determination unit and re-encoding the image by referring to the bandwidth (bit rate) and image quality (distortion) information received from the rate-distortion measurement unit. And a video code for transmitting the recoded video and the compressed stream to the rate-distortion measurement unit. The rate-distortion measurement unit 400 measures the image quality of the re-encoded image received from the image encoder and the bandwidth (bit rate) of the compressed stream, and transmits the measured image quality and bandwidth (bit rate) to the image encoder. It is characterized by consisting of. In addition, the compressed stream which is the input information of the image decoder 100 will be described by using an example of a re-encoding method encoded by the H.265 / HEVC compression standard. Table 1 below compares the H.264 / AVC compression standard and H.265 / HEVC coding tools.

TABLE 1

For reference, the H.264 / AVC compression standard and the H.265 / HEVC compression standard are compression standards created by the collaboration of both ISO MPEG and ITU-T VCEG, which are widely used in the image-based product and service market. Referring to Table 1, it can be seen that the H.265 / HEVC compression standard provides a wider variety of encoding tools than the H.264 / AVC compression standard. Compared to the .264 / AVC compression standard, it can provide a good compression ratio of about 30% to 50%.

4 is a control flowchart of a decoding information based image recoding method of a camera image of the present invention. In FIG. 4, in the decoding information-based image re-coding method of the camera image of the present invention, a step of decoding an entire frame with respect to the compressed stream inputted by the image decoder (S11), and the decoding information output by the image decoder is transmitted to the re-coding determination unit. (S12) to determine the re-encoding method, and transmitting the re-encoding method determined by the re-encoding determination unit to the image encoder (S13), and the re-encoding that the image encoder receives and receives the reconstructed image from the image decoder. Based on the re-coding received from the rate-distortion measurement unit and the image quality of the image and the bandwidth (bit rate) of the compressed stream, the prediction unit and the conversion unit encoding method of the 64x64 coding unit of the entire frame are based on the rate-distortion optimization technique. Determining, encoding, and outputting the image (S14), and after re-encoding the image, Comparing the first bit rate with the target bit rate, measuring the image quality of each 64x64 block of the frame corresponding to the case where the actual bit rate is smaller than the target bit rate, and repeatedly performing image re-encoding of the low quality region (S15); And if the bit rate reaches the target bit rate, terminating the video re-encoding (S16). The compressed stream input in step S11 is encoded by the H.264 / AVC compression standard, and the output of the image encoder is encoded by the H.265 / HEVC compression standard in step S14. In addition, in step S12, the H.264 / AVC compression standard has a fixed coding unit of 16x16, whereas the H.265 / HEVC compression standard can have a coding unit having a maximum size of 64x64. Since the amount of bits required for display is small, the 64x64 coding unit can be determined as the initial coding information. Therefore, the re-encoding determiner determines that the encoded information received from the image decoder 100 is 16x16 size, whereas the initial encoded information is determined to be 64x64 size. In addition, the integrated determination may be integrated determination of quantization coefficient information for quantizing the motion vector for the motion estimation and the residual signal. The integrated determination method of the motion vector and the quantum and the coefficient as described above

(One). An initial motion vector of a 64x64 coding unit is calculated as in Equation (1) below.

……(식 1)

… … (Equation 1)

In Equation (1), MVCTU denotes an initial motion vector of a 64x64 block, and MVPixe (y, x) denotes a motion vector of a pixel corresponding to the vertically y th and horizontal x th positions inside the 64 x 64 block. . Initial motion vector The vertical and horizontal elements of the MVCTU are independently determined by the average value of the vertical and horizontal elements of MVPixe (y, x), and MVPixe (y, x) is the motion vector of the block that contains the pixel. It means.

(2). Since the larger the quantization coefficient value, the smaller the bit amount is required to encode the residual signal, the initial quantization coefficient information determines the quantization coefficient value as the maximum value among the quantization values of 16 16x16 blocks included in the 64x64 block.

(2). The initial coding unit is 64x64, the initial motion vector is MVCTU, and the recoding method based on the determined initial quantization coefficient value is transmitted to the image encoder.

Further, in the step S14, the rate-distortion optimization technique is a search process for finding a method of encoding with an optimal image quality based on the amount of bits used, and encodes a method of minimizing the rate-distortion cost calculated by Equation (2) below. How to decide.

……(식 2)

… … (Equation 2)

Distortion in Equation (2) means a distortion value between the pixel value of the original image input to the encoder (in the present invention, a reconstructed image output from the image decoder) and the pixel value compressed by the encoding method. It is calculated together.

……(식 3)

… … (Equation 3)

In Equation (3), orgImg (y, x) and recImg (y, x) mean pixel values of the original image and the image compressed by the encoding method, respectively. In addition, the rate refers to the amount of bits required for encoding by the method, and λ is a Lagrangian parameter. The motion estimation process for removing temporal overlapping components of an image signal during image encoding requires a large amount of image computation and increases the complexity of the encoder. In the present invention, since the initial motion vector is selected by using the existing decoding information, since a separate motion estimation process is not performed, encoding can be performed at high speed by reducing the time required for image recoding.

In addition, the image quality measurement for determining the low quality region in step S15 refers to the distortion value of the 64x64 block calculated by Equation (3). In other words, the larger the distortion value, the lower the quality. On the contrary, the smaller the distortion value, the higher the quality.

5 is a control flowchart of a re-coding method of a low quality region applied to the present invention. In FIG. 5, the re-encoding method of the low quality area is to increase the quality of the corresponding area after determining the low quality area in step S15. The method includes re-encoding the motion vector by improving the motion vector (S31), and whether the corresponding area is the low quality area. Determining (S32), and if the corresponding area is a low quality area, then re-encoding by improving the quantization coefficient value (S33), determining whether the area is a low quality area (S34), and In the case where the region is a low quality region, a next step of reducing the coding unit of the region and recoding the region is performed. In addition, in the step S31, the improvement of the motion vector is performed by resetting the motion region by setting the peripheral ± 4 pixel range as the search region around the initial motion vector MVCTU, thereby finding a more accurate motion vector than the initial motion vector MVCTU. This is to improve the picture quality. In addition, to improve the quantization coefficient value in the step S33 is to determine the optimal quantization coefficient value by performing rate-distortion optimization while reducing the initial quantization coefficient value by -1 three times. In addition, in step S35, since the coding unit of the corresponding region is reduced by 64x64 coding units during initial encoding, the coding units are performed in the order of 32x32 and 16x16 during recoding. The quality of the image can be improved by re-coding each of them using an encoding method most suitable for the corresponding region. Also. If it is determined in S31 or S33 that the corresponding region is not a low quality region, the recoding is terminated.

100: video decoding unit, 200: re-encoding determination unit,
300: image encoder, 400: rate-distortion measurement unit

Claims (14)

delete delete delete In the video re-encoding method based on the decoding information of the camera image,
Decoding information-based image re-coding method of the camera image,
Decoding an entire frame with respect to the compressed stream inputted by the image decoder (S11);
Transmitting decoded information output from the image decoder to the re-encoding determination unit to determine a re-encoding method (S12);
Since the encoding information transmitted by the image decoding unit is 16x16 in size, the initial encoding information is determined as 64x64 in size, and thus the recoding determination unit combines and determines the motion vector for motion estimation and the quantization coefficient information for quantizing the residual signal. Transmitting to the image encoder (S13);
Based on the rate-distortion optimization technique based on the quality of the re-encoded image received from the image decoder and the rate-distortion measurement unit and the bandwidth of the compressed stream, the image encoder receives the reconstructed image from the image decoder. Determining, encoding, and outputting a prediction unit and a transformation unit encoding method of a 64x64 coding unit (S14);
After re-encoding the image, comparing the actual bit rate required for the corresponding frame encoding with the target bit rate, after measuring the image quality of each 64x64 block of the frame corresponding to the case where the actual bit rate is smaller than the target bit rate, the image re-encoding of the low quality region is repeatedly performed. Step S15;
And terminating image re-encoding when the actual bit rate of the frame reaches the target bit rate (S16). The initial motion vector of the 64x64 coding unit includes:

The video re-encoding method based on the decoding information of the camera image, characterized in that calculated by.
MVCTU is the initial motion vector of the 64x64 block, and MVPixe (y, x) is the motion vector of the pixel corresponding to the y-th vertically and x-th horizontally within the 64x64 block.
The method of claim 4, wherein
The input compressed stream of the image decoder is
Encoded in the H.264 / AVC compression standard,
The output of the video encoder,
Coded with the H.265 / HEVC compression standard, the H.264 / AVC compression standard is fixed to 16x16 coding unit, the H.265 / HEVC compression standard can have a coding unit of up to 64x64 size An image re-encoding method based on decoding information of an image.
delete The method according to claim 4 or 5,
The optimization technique of the step S14,

And a coding method for determining that the rate-distortion cost calculated by the minimum value is minimized.
Here, Distortion is a distortion value between the pixel value of the original image input to the encoding unit (in the present invention, the reconstructed image output from the image decoding unit) and the pixel value compressed by the encoding method, and Rates is the amount of bits required when encoding by the corresponding method Λ is Lagrangian parameter.
The method according to claim 4 or 5,
In the step S15, the image quality measurement for determining the low quality region,

Decoding information-based image re-encoding method for a camera image, characterized in that referring to the distortion value of the 64x64 block calculated by.
Where orgImg (y, x) and recImg (y, x) are the pixel values of the original image and the image compressed by the encoding method, respectively.
The method according to claim 4 or 5,
The video re-coding method of the low quality region,
Re-encoding by performing motion vector improvement;
Determining whether the corresponding area is a low quality area (S32);
If the corresponding area is a low quality area, re-encoding the quantization coefficient value next (S33);
Determining whether the corresponding area is a low quality area (S34);
And re-encoding by reducing the coding unit of the corresponding area when the corresponding area is a low quality area (S35).
delete The method according to claim 4 or 5,
The initial quantization coefficient information is
A decoding information-based image re-encoding method of a camera image, characterized by determining a quantization coefficient value as a maximum value among quantization values of 16 16x16 blocks included in a 64x64 block.
The method of claim 9,
The improvement of the motion vector in step S31,
A decoding information-based image re-encoding method of a camera image, comprising: performing a motion region again by setting a peripheral ± 4 pixel range as a search region around the initial motion vector MVCTU.
The method of claim 9,
Improving the quantization coefficient value in the step S33,
Decoding information-based image re-encoding method for a camera image, characterized in that the optimal quantization coefficient value is determined by performing rate-distortion optimization while reducing the initial quantization coefficient value by -1 three times.
The method of claim 9,
In operation S35, reducing the coding unit of the corresponding region may be small.
Since the 64x64 coding units are performed during the initial encoding, the coding units are performed in the order of 32x32 and 16x16 during recoding. An image re-encoding method based on decoding information of a camera image.

Images