A live scene adaptive video encoding method

Abstract

The adaptive video encoding method for live broadcast scenarios of the present invention reduces the resolution and bit rate by splitting the image into four channels of video at half the resolution, while simultaneously stitching the four channels back together to create a high-resolution, high-quality video, thereby adapting to different network bandwidths and devices. Compared to traditional methods, the present invention is equivalent to encoding only one channel of high-resolution video, which is simply split to achieve both low and high resolutions. While the coding efficiency, network bandwidth usage, and storage usage are equivalent to that of a single channel of high-resolution video in the traditional method, different resolutions and bit rates are achieved to meet the needs of different devices and network bandwidth conditions.

Claims

1. A live scene adaptive video encoding method, characterized in that: It includes the following steps: S0. Read a frame of original image data from the camera; S1. Divide the image pixel points read in step S0 into groups of four pixels in a cross shape with two pixels each above and below, and obtain multiple four-pixel groups in sequence; S2. For each four-pixel group divided in step S1, extract the pixel in the upper left corner, and form a new image A with these pixels; S3. According to the method of step S2, extract and form new images B, C, and D with the pixels in the upper right corner, lower left corner, and lower right corner of each four-pixel group respectively. The original image is divided into four complete images A, B, C, and D with a resolution of one-fourth; S4. Add the corresponding pixel points of images A, B, C, and D, and then calculate the average value to obtain an averaged image A'; S5. Encode the averaged image A' obtained in step S4, and record the eight prediction methods with the highest intra-frame prediction scores during the encoding process; S6. Encode the images B, C, and D obtained in step S3, and perform prediction using the intra-frame prediction methods recorded in step S5; S7. Repeat steps S0 to S6 to process and encode the next frame of image, and obtain four encoded video raw bitstreams 1, 2, 3, and 4; S8. Multiplex and encapsulate the video raw bitstream 1 obtained in S7 and the audio data into a transport stream; S9. Directly encapsulate the other three raw bitstreams 2, 3, and 4 into the transport stream; S10. Transmit the four transport streams respectively through the streaming media transmission protocol.

2. The live scene adaptive video encoding method according to claim 1, characterized in that: In step S0, the live broadcast device reads a frame of collected original video data in YUV format from the camera.

3. The live scene adaptive video encoding method according to claim 1, characterized in that: In step S1, divide the frame of YUV data obtained in step S0. Divide the image pixel points into a cross shape with four pixels of two pixels each above and below in sequence. If the number of pixels in the horizontal or vertical direction is not an integer multiple of 2, discard the rightmost column or the bottom row of pixels to ensure that the image can be completely divided into a cross shape.

4. The live scene adaptive video encoding method according to claim 1, characterized in that In step S2, for each four-pixel group obtained by division in step S1, extract the pixel in the upper left corner of the cross, and form a new image A with these pixels in the original order. The new image A is complete compared to the original image, but the resolution is only one-fourth of the original image.

5. The live scene adaptive video encoding method according to claim 1, characterized in that: In step S4, average the four images A, B, C, and D obtained in step S3. Take the corresponding pixels in images A, B, C, and D in sequence, add the Y values of the four pixel YUV data and divide by four to obtain the average value of Y. Process the U and V values of the four pixel YUV data in the same way to obtain the average value of the four pixel YUV. Then form all the YUV average values of the four images into an averaged image A'.

6. The live scene adaptive video encoding method according to claim 1, characterized in that In step S5, encode the averaged image A' obtained in step S4 using the third-generation audio and video standard (AVS3) encoding format, and record the eight prediction methods with the top eight intra-frame prediction scores during the encoding process for intra-frame prediction of the other three images.

7. The live scene adaptive video encoding method according to claim 1, characterized in that: In step S6, the images B, C, and D obtained in S3 are encoded and predicted using the eight prediction methods recorded in S5. Since the contents of the images B, C, and D are the same as the averaged image A', their intra-frame prediction methods are the same. There is no need to traverse all the prediction methods, and only the prediction method recorded in step S5 is used for prediction.

8. The live scene adaptive video encoding method according to claim 1, characterized in that: In step S8, the video original stream 1 obtained in S7 and the audio data collected by the microphone are multiplexed, and the decoding time stamp (DTS) and presentation time stamp (PTS) information are added, and then encapsulated into a transmission stream.

Images