Image Quality Processing via Low-Resolution Model Upscaling
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The increase in high-resolution content processing in electronic devices leads to issues such as increased complexity, costs, and power consumption due to the need for larger IP blocks for image quality processing.
Innovation Solution
Down-sample high-resolution input images to obtain low-resolution images, model the conversion relationship between these images before and after processing, and use this model to derive high-resolution models for efficient image quality processing.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If the resolution of content input to the electronic devices is increased, then the image quality processing capability is improved, but the size of IP block needs to be increased which leads to increased complexity, costs, power consumption, and heat generation
Solution Approach 1:
The patent segments the image processing task by first down-sampling the high-resolution input image to a low-resolution image, processing this reduced-size image through the IP block, and then up-sampling the result back to high-resolution. This segmentation allows the IP block to operate at lower resolution (reducing its size requirements) while still handling high-resolution content through the combined pipeline of down-sampling, processing, and up-sampling operations.
Solution Approach 2:
The patent introduces an intermediary approach by using a low-resolution representation of the image as a mediator between the high-resolution input and the final high-resolution output. The low-resolution image serves as an intermediate form that the IP block processes, enabling the system to achieve high-resolution output without requiring the IP block itself to be sized for high-resolution input directly.
2Measurement precision
If the size of the IP block is increased to handle high-resolution content, then the image quality processing capability is improved, but power consumption increases
Solution Approach 1:
The patent segments the processing workload by operating the IP block on down-sampled low-resolution images rather than full high-resolution images. This segmentation reduces the computational burden on the IP block, thereby reducing power consumption while still achieving high-resolution output through the subsequent up-sampling step.
Solution Approach 2:
The patent applies partial action by processing only the essential low-resolution representation of the image through the IP block, rather than processing every high-resolution pixel. The down-sampling step selects key information at lower resolution, and the up-sampling step reconstructs the high-resolution output, achieving the necessary processing with reduced energy expenditure.
3Measurement precision
If the size of the IP block is increased to handle high-resolution content, then the image quality processing capability is improved, but heat generation increases
Solution Approach 1:
The patent segments the processing task to operate the IP block on low-resolution images obtained through down-sampling. This segmentation reduces the computational load and thermal generation within the IP block, as fewer pixels require processing. The final high-resolution output is achieved through up-sampling without requiring the IP block to handle the full high-resolution data load directly.
4Device complexity
If down-sampling is applied to reduce image resolution for processing, then the IP block size and complexity are reduced, but image quality may be lost
Solution Approach 1:
The patent incorporates feedback through the up-sampling step that takes the low-resolution processed image and reconstructs it to high-resolution. This feedback mechanism allows the system to compensate for the information loss introduced by down-sampling, using the processed low-resolution image as a guide to regenerate the high-resolution output with improved quality compared to the original input.
Solution Approach 2:
The patent changes the resolution parameter dynamically in the processing pipeline: the input image is down-sampled to a lower resolution for IP block processing, and then the output is up-sampled back to high-resolution. This parameter change allows the IP block to operate efficiently at lower resolution while the system maintains high-resolution capability through the transformation pipeline.
Data Source
AI summary
A method of an electronic device including obtaining a low-resolution input image by down-sampling a high-resolution input image; obtaining a low-resolution output image by performing image quality processing on the low-resolution input image; obtaining a low-resolution model from a conversion relationship between the low-resolution input image prior to the image quality processing being performed and the low-resolution output image subsequent to the image quality processing being performed; performing up-sampling of the low-resolution model; obtaining a high-resolution model by modifying the up-sampled low-resolution model, based on a difference between the high-resolution input image and the low-resolution input image; and obtaining a high-resolution output image from the high-resolution input image, by applying the high-resolution model to the high-resolution input image.