AI Image Generation Prompt Refinement for Faster High-Resolution Output
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing electronic devices face significant delays in generating and providing high-resolution images due to the time-consuming process of upscaling images received from external servers, degrading user experience and requiring repetitive processes if the generated image is not accepted.
Innovation Solution
An electronic device modifies user input prompts based on feature information to generate candidate images of varying quality, selects a preferred image, and enhances the quality of the final image through neural networks, either locally or via an external server, to expedite the delivery of high-quality images.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Manufacturing precision
If the electronic device performs upscaling processing locally after receiving images from the external server, then the image quality is improved, but the processing time and total wait time increase significantly
Solution Approach 1:
The image processing pipeline is segmented into two distinct stages: (1) a fast generative AI model on the external server that produces initial candidate images of acceptable quality, and (2) a slower high-quality generative AI model that refines the selected candidate image. This segmentation allows the system to achieve high image quality without requiring the slow model to process all images from scratch, thereby reducing total processing time while maintaining quality.
Solution Approach 2:
The external server performs preliminary image generation using a fast generative AI model to create multiple candidate images before the user makes a selection. This preliminary action provides the user with immediate visual feedback and allows the system to prepare multiple options in advance, reducing the overall wait time when the user eventually selects an image for high-quality processing.
2Device complexity
If the electronic device transmits images to the external server for processing, then the device complexity is reduced, but the dependency on external services increases
Solution Approach 1:
The electronic device is designed with multi-functionality to handle various image processing tasks: it can locally execute fast generative AI models for initial image generation, perform quality assessment to evaluate candidate images, execute high-quality generative AI models for refinement when needed, and communicate with external servers when necessary. This universal design reduces device complexity for specific tasks while maintaining adaptability across different processing scenarios.
Solution Approach 2:
The electronic device performs self-service by autonomously executing local generative AI models and quality assessment algorithms without requiring constant external server intervention. The device can independently generate candidate images, evaluate their quality, and only resort to external servers or high-quality processing when necessary, thereby reducing dependency on external services while maintaining processing capabilities.
3Ease of operation
If the electronic device displays multiple candidate images for user selection, then the user experience is improved, but the processing time and resource consumption increase
Solution Approach 1:
The system generates and displays a limited number of high-quality candidate images (e.g., top 3-5) rather than all possible generated images. By using quality assessment to filter and select only the most promising candidates for display, the system provides sufficient user experience improvements while significantly reducing processing time and computational resources compared to generating and displaying all possible images.
Data Source
AI summary
An electronic device includes memory storing instructions; and at least one processor, wherein the instructions, when executed, cause the electronic device to receive an input prompt; identify feature information in a first portion of a first generated image based on the input prompt; obtain a modified prompt by modifying the input prompt based on the feature information; obtain candidate images of a first image quality corresponding to the modified prompt using a first generative artificial intelligence (AI) model; display a user interface (UI) including the candidate images via a display; and based on a selected candidate image being identified from among the candidate images, obtain a second generated image of a second image quality corresponding to the selected candidate image, wherein a second image quality parameter of the second generated image is higher than a first image quality parameter of the first generated image.