Video Processing Apparatus Using DRSCN Model for Bandwidth Reduction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
High-definition videos with high frame rates require substantial bandwidth for transmission, exceeding typical transmission capacities, and existing video codecs struggle to balance compression ratio and video quality, particularly for high-resolution and high-frame-rate videos.
Innovation Solution
A video processing apparatus and method that includes a down-scaling module, a learning machine module to generate a trained model, and a transmitter to compress both the down-scaled video file and the model, using a deep residual convolution neural network (DRSCN) for data reduction, and a decompressor to reconstruct the original video file.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Quantity of substance
If video compression is applied to reduce bandwidth requirements, then transmission bandwidth is reduced, but video quality deteriorates
Solution Approach 1:
The patent segments the video processing into multiple stages: down-scaling to reduce resolution, compression to reduce data volume, and up-scaling with neural network enhancement to restore quality. This multi-stage segmentation allows achieving both data reduction and quality preservation that single-stage compression cannot accomplish
Solution Approach 2:
The patent introduces a neural network model as an intermediary between compression and decompression. This trained model learns the mapping relationship between compressed and original videos, acting as a mediator that reconstructs high-quality video from compressed data, thereby resolving the quality loss caused by compression
2Quantity of substance
If high compression ratio is used to reduce bandwidth, then bandwidth requirements are reduced, but video quality becomes unacceptable
Solution Approach 1:
The patent performs preliminary down-scaling of the video before compression. By reducing the video resolution in advance, the total data volume is significantly reduced, allowing for more effective compression while maintaining acceptable quality after the neural network reconstruction process
Solution Approach 2:
The patent changes key parameters of the video processing system: it transforms the video from high resolution to low resolution temporarily for compression, then uses neural network enhancement to restore quality. It also changes the compression approach from traditional lossy compression to a combination of down-scaling compression and neural network-based reconstruction
3Manufacturing precision
If traditional video codecs are used for high-resolution high-frame-rate videos, then video quality is maintained, but bandwidth requirements exceed transmission capacity
Solution Approach 1:
The patent replaces traditional mechanical video compression codecs with a neural network-based compression and reconstruction system. This substitution enables more efficient data representation and quality restoration, achieving superior performance in both bandwidth reduction and quality preservation compared to conventional codecs
Data Source
AI summary
The disclosure is directed to a video processing apparatus and a video processing method thereof. In one of the exemplary embodiments, the disclosure is directed to a video processing apparatus which includes not limited to a storage medium configured to store a first video file, a down-scaling module coupled to the storage medium and configured to down-scale the first video file into a second video file, a learning machine module configured to receive the first video file and a third video file which is processed from the second multimedia file and generate a trained model out of the first video file and the third video file, and a transmitter configured to transmit a data package which comprises a compression of the second video file and a compression of the trained model.