Neural Network Image Transmission System Bandwidth Compression
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Video image content transmission over limited bandwidth networks faces challenges with high data volume, leading to increased communication path load and potential congestion, and existing compression methods often result in decreased image quality with block noise and mosquito noise.
Innovation Solution
An image transmission and reception system utilizing machine learning to generate model data from low-bitrate-encoded images, which are then transmitted along with the encoded images, allowing the reception device to restore improved images with quality similar to the original using machine-learned model data.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Manufacturing precision
If video image content is transmitted over limited bandwidth networks, then image quality can be maintained, but communication path load increases and congestion occurs
Solution Approach 1:
The video stream is divided into multiple scenes, and each scene is further segmented into blocks. This segmentation allows selective transmission of only necessary block data and differential data, significantly reducing the overall data volume while maintaining image quality for important regions.
Solution Approach 2:
The patent extracts and transmits only the essential components: scene type identification data, block priority data, and differential data for selected blocks. By taking out only the necessary information rather than transmitting complete high-quality video data, the communication load is reduced while preserving image quality.
2Quantity of substance
If data transmission rate is reduced to compress bandwidth, then transmission load decreases, but image quality deteriorates with block noise and mosquito noise
Solution Approach 1:
Different blocks within the same video frame are assigned different priorities based on their importance. High-priority blocks (containing important visual information) are transmitted with higher quality or fully transmitted, while low-priority blocks use more aggressive compression. This local differentiation maintains overall image quality while reducing total data transmission.
Solution Approach 2:
Instead of uniformly compressing all video data, the patent applies partial action by selectively transmitting differential data only for specific blocks that need it. For high-priority blocks, full data is transmitted, while for low-priority blocks, only essential differential information is sent, achieving quality preservation where needed and compression where acceptable.
Data Source
AI summary
In a system for distributing video image content from a server to a viewer terminal, a configuration is provided in which the load to the transmission path is reduced by reducing the volume of distribution volume while improving the image quality to be viewed. Content data based on a low-bitrate-encoded image and data of an image of a transformation matrix in a neural network which is model data for obtaining an image close to an original image from the low-bitrate-encoded image are transmitted to the viewer terminal from the video image content distribution server. These data are used by the viewer terminal to obtain improved video image content.