用于基于神经网络的视频译码的前端架构
By processing YUV format video data through a neural network system based on end-to-end machine learning, generating a combined representation and performing quantization and entropy coding, the problem of low video decoding efficiency in existing technologies is solved, and efficient video data compression is achieved.
CN116547965BActive Publication Date: 2026-07-17QUALCOMM INC
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- QUALCOMM INC
- Filing Date
- 2021-12-09
- Publication Date
- 2026-07-17
AI Technical Summary
Technical Problem
Existing video decoding technologies struggle to effectively compress video data while maintaining high quality, leading to excessive burdens on communication networks and equipment.
Method used
A neural network system based on end-to-end machine learning, especially a convolutional neural network, is used to process YUV format video data. It generates a combined representation through convolutional and nonlinear layers, and performs quantization and entropy coding to adapt to YUV 4:2:0 format video decoding.
Benefits of technology
It improves the efficiency and quality of video decoding, reduces the amount of data, and lowers the processing and storage burden on devices.
✦ Generated by Eureka AI based on patent content.
Smart Images
Figure CN116547965B_ABST
Abstract
本文描述了用于使用神经网络系统来处理视频数据的技术。例如，一种过程可以包括：由神经网络系统的编码器子网络的第一卷积层生成与帧的亮度通道相关联的输出值。该过程可以包括：由编码器子网络的第二卷积层生成与帧的至少一个色度通道相关联的输出值。该过程可以包括：由第三卷积层基于与帧的亮度通道相关联的输出值和与帧的至少一个色度通道相关联的输出值来生成帧的组合表示。该过程可以包括：基于帧的组合表示来生成经编码的视频数据。
Need to check novelty before this filing date? Find Prior Art