AI Image Upscaling With Server Downscaling for Limited Bandwidth
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The challenge of transmitting high-definition images in limited network resources is exacerbated by the mismatch between increasing video capacity and stagnant network bandwidth, necessitating improved image compression and restoration techniques.
Innovation Solution
An electronic apparatus and server system that utilizes a Convolutional Neural Network (CNN) to downscale and upscale image data using a trained filter set, minimizing differences between original and restored images through selective filter application.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If image data is transmitted at high definition quality, then image quality is improved, but network resource consumption increases
Solution Approach 1:
The image processing is segmented into two stages: downscaling at the server端 to reduce transmission data量, and upscaling at the terminal端 to restore image quality. This segmentation allows the system to transmit compressed image data while maintaining high definition output quality, resolving the contradiction between image quality and network resource consumption
Solution Approach 2:
The server performs preliminary downscaling and compression of image data before transmission. By pre-processing the image data to reduce its size while preserving essential visual information, the system minimizes network resource consumption during transmission while enabling high-quality reconstruction at the terminal端
2Quantity of substance
If image data is compressed to reduce transmission size, then network resource consumption is reduced, but image quality deteriorates
Solution Approach 1:
Traditional mechanical compression algorithms are replaced with AI-based neural network models for both downscaling and upscaling operations. The AI models learn optimal compression and reconstruction strategies, achieving superior image quality preservation compared to conventional compression methods while maintaining reduced transmission sizes
Solution Approach 2:
The system changes the parameter of image resolution dynamically by downsampling at the server端 and upsampling at the terminal端. This parameter transformation allows the same image data to be transmitted in a compressed form while being reconstructed at higher resolution, effectively decoupling transmission size from final image quality
3Measurement precision
If AI models with more filters are used for upscaling, then image quality is improved, but device complexity increases
Solution Approach 1:
The system shifts the computational complexity from the terminal端 to the server端 by performing the complex downscaling operation with multiple filters at the server. The terminal端 only needs to execute the upscaling operation, which can use fewer filters since the heavy compression work has already been done. This dimensional shift in where complexity is applied resolves the contradiction between image quality and device complexity
Data Source
AI summary
The disclosure relates to an artificial intelligence (AI) system that uses a machine learning algorithm and an application thereof. A method for controlling an electronic apparatus according to the disclosure includes receiving image data and information associated with a filter set that is applied to an artificial intelligence model for upscaling the image data from an external server; decoding the image data; upscaling the decoded image data using a first artificial intelligence model that is obtained based on the information associated with the filter set; and providing the upscaled image data for output.