AI Metadata Generation for Image Quality and Load Reduction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing image reproducing apparatuses face challenges in maintaining image quality during the transmission and reconstruction of digital images due to data encoding and processing, which often results in deteriorated image quality and increased processing load.
Innovation Solution
A lightweight artificial intelligence (AI) network is employed to generate AI metadata, including class information and class maps, to improve image quality by encoding and decoding images while reducing processing resources and time, using a first AI network for image feature extraction and a second AI network for image-quality enhancement.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Manufacturing precision
If image processing is performed by extracting image features from the input image to improve image quality, then image quality is improved, but processing load increases and processing speed decreases
Solution Approach 1:
The patent applies preliminary action by generating AI metadata from the original image before encoding and transmission. This metadata contains pre-extracted feature information (object types, regions, attributes) that can be directly used during reproduction without requiring complex feature extraction at the reproducing apparatus. The feature extraction is performed in advance at the source apparatus, transferring the computational burden from the reproducing side to the transmission side.
Solution Approach 2:
The patent introduces AI metadata as an intermediary between the original image and the reproduced image. This metadata acts as a bridge that carries essential feature information through the transmission channel, enabling the reproducing apparatus to reconstruct image quality without performing complex processing. The metadata includes class information, class maps, and attribute information that mediate the reconstruction process.
2Manufacturing precision
If image processing is performed by extracting image features from the input image to improve image quality, then image quality is improved, but processing speed decreases
Solution Approach 1:
Feature extraction and analysis are performed in advance at the image providing apparatus before encoding. The AI metadata containing all necessary feature information is generated and attached to the encoded image data, eliminating the need for time-consuming feature extraction at the reproducing apparatus during real-time playback or reproduction.
Solution Approach 2:
The patent extracts only the essential feature information (object classes, regions, attributes) from the full image data and places it into a compact metadata structure. This selective extraction of critical features reduces the amount of data that needs to be processed during reproduction, significantly improving processing speed while maintaining image quality enhancement capabilities.
3Manufacturing precision
If AI metadata is generated from the original image using a first AI network, then image quality enhancement is enabled, but data transmission volume increases
Solution Approach 1:
The patent extracts only the most essential feature information from the original image and encodes it into a compact metadata format. Instead of transmitting the entire original image or all possible feature data, only critical elements (class information, class maps, key attributes) are included in the AI metadata, minimizing the additional data volume while maintaining image quality enhancement capability.
Solution Approach 2:
The patent transforms rich, detailed image feature data into a parameterized metadata structure with defined schemas and compression formats. By converting complex image data into structured parameters (class labels, region coordinates, attribute values), the data is significantly compressed while retaining the essential information needed for quality enhancement during reproduction.
Data Source
AI summary
An image providing apparatus configured to generate, by using a first artificial intelligence (AI) network, AI metadata including class information and at least one class map, in which the class information includes at least one class corresponding to a type of an object among a plurality of predefined objects included in a first image and the at least one class map indicates a region corresponding to each class in the first image, generate an encoded image by encoding the first image, and output the encoded image and the AI metadata through the output interface.