AI Image Encoding and Decoding via Joint DNN Training
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current image processing technologies face challenges in efficiently encoding and decoding high-resolution images, leading to increased bitrate and reduced image quality, especially when handling high-definition images.
Innovation Solution
The implementation of artificial intelligence (AI) encoding and decoding methods that involve joint training of deep neural networks (DNNs) for down-scaling and up-scaling, optimizing DNN setting information to reduce the amount of information needed for encoding and decoding, and using techniques like differential pulse code modulation, run-length coding, and Huffman coding for efficient data transmission.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Manufacturing precision
If conventional encoding methods are used for high-resolution images, then image quality can be maintained, but bitrate increases significantly
Solution Approach 1:
The patent replaces conventional mechanical encoding/decoding systems with AI-based neural network systems. The encoder uses a down-scaling DNN to convert high-resolution images to low-resolution representations, while the decoder uses an up-scaling DNN to reconstruct high-resolution images from these compressed representations, achieving both quality preservation and bitrate reduction
Solution Approach 2:
The patent changes the fundamental parameters of image encoding by transforming spatial domain operations into learned transformations through neural networks. The DNN models learn optimal parameter transformations during training, enabling efficient representation of high-resolution images at lower bitrates while maintaining manufacturing precision
2Manufacturing precision
If DNN setting information is frequently updated for AI up-scaling, then image quality improves, but the amount of information to be encoded increases
Solution Approach 1:
The patent applies partial updates to DNN setting information rather than complete re-encoding. The system updates only necessary portions of the neural network parameters (such as scaling factors or specific layer configurations) based on current image characteristics, avoiding the need to encode entire DNN models while still improving image quality when needed
Solution Approach 2:
The patent performs preliminary training of the up-scaling DNN using joint training with the down-scaling DNN on representative image data. This pre-training establishes baseline DNN setting information that can be stored and reused, reducing the need for frequent updates and minimizing the amount of information that needs to be encoded and transmitted
Data Source
AI summary
An artificial intelligence (AI) decoding method including obtaining image data generated from performing first encoding on a first image and AI data related to AI down-scaling of at least one original image related to the first image; obtaining a second image corresponding to the first image by performing first decoding on the image data; obtaining, based on the AI data, deep neural network (DNN) setting information for performing AI up-scaling of the second image; and generating a third image by performing the AI up-scaling on the second image via an up-scaling DNN operating according to the obtained DNN setting information. The DNN setting information is DNN information updated for performing the AI up-scaling of at least one second image via joint training of the up-scaling DNN and a down-scaling DNN used for the AI down-scaling.