AI Image Scaling for Low-Bitrate Display Window Rendering
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The increasing demand for high-resolution and high-quality image encoding and decoding has led to challenges in managing storage capacity and bitrate, particularly in hardware capable of reproducing and storing such images, necessitating more efficient encoding and decoding methods.
Innovation Solution
An AI-based approach is employed for image encoding and decoding, utilizing neural networks for down-scaling and up-scaling operations to manage storage capacity and bitrate effectively, with processes involving AI down-scaling, first encoding, and first decoding, followed by AI up-scaling to achieve optimal image quality and size for display.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Manufacturing precision
If high-resolution and high-quality image encoding is performed, then image quality is improved, but storage capacity is consumed faster and bitrate increases
Solution Approach 1:
The patent applies AI-based dynamic parameter adjustment in the encoding process, where neural networks analyze image content and adaptively modify encoding parameters such as quantization levels, bitrate allocation, and resolution scaling factors. This allows high-quality encoding for important image regions while using lower bitrates for less critical areas, thereby improving overall image quality without proportionally increasing storage capacity consumption
Solution Approach 2:
The patent implements region-based differential encoding where different parts of the image are encoded with different quality levels and bitrates based on their importance. AI algorithms identify salient regions and allocate more bits to these areas while using compression for background or less important regions, achieving high perceived quality at lower overall storage requirements
2Manufacturing precision
If high-resolution and high-quality image encoding is performed, then image quality is improved, but bitrate increases
Solution Approach 1:
The patent uses AI-driven dynamic parameter optimization where neural networks continuously adjust encoding parameters including quantization matrices, transform block sizes, and prediction modes based on local image characteristics. This adaptive approach maintains high image quality in critical regions while minimizing bitrate in less important areas, resolving the contradiction between quality and bitrate
Solution Approach 2:
The patent implements dynamic bitrate allocation and resolution scaling where the encoding parameters are not fixed but continuously adapted based on scene complexity, motion, and importance maps generated by AI models. This allows the system to maintain high quality when needed while reducing bitrate during static or less important scenes, achieving quality-bitrate optimization
3Quantity of substance
If AI-based encoding and decoding is performed, then storage capacity is saved and bitrate is reduced, but processing complexity increases
Solution Approach 1:
The patent divides the complex AI processing into separate modules: an AI encoder for compression, a conventional decoder for basic decoding, and an AI upscaler for quality enhancement. This segmentation allows each component to be optimized independently and processed in parallel where possible, reducing the computational burden on any single processor while achieving overall storage and bitrate efficiency
Solution Approach 2:
The patent introduces an intermediary AI upscaling stage between the conventional decoder and final display. The conventional decoder handles the computationally intensive decoding of compressed data, while the lighter-weight AI upscaler enhances quality. This intermediary approach separates the heavy lifting from the enhancement, managing processing complexity more effectively
Data Source
AI summary
Provided is a display apparatus including a processor configured to execute instructions to identify whether a down-scale ratio between a size of an image reconstructed from image data of a previous image and a size of a first window is equal to or less than a pre-set limiting down-scale ratio, transmit a request for a first current image to an image providing apparatus, based on a result of the identifying, and display, on the first window, a second current image or a down-scaled second current image by performing first decoding on image data of the first current image received from the image providing apparatus, wherein a ratio between a size of the second current image and the size of the first window is equal to or greater than the pre-set limiting down-scale ratio.