AI Region of Interest Extraction for High-Resolution Image Processing
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing electronic devices face challenges in processing high-resolution images efficiently due to the high computational requirements and memory needs, leading to delays in generating and processing high-resolution content with limited resources.
Innovation Solution
An electronic apparatus and method that downscale high-resolution images to a critical resolution, using a first artificial intelligence model to identify regions of interest and a second model to analyze these regions, thereby reducing computational load and memory requirements, allowing for efficient processing and feature extraction of specific object regions.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If high-resolution images are processed directly by artificial intelligence models, then object recognition accuracy is improved, but computational load and processing time increase significantly
Solution Approach 1:
The patent applies segmentation by dividing the high-resolution image processing into two stages: first processing a downsampled version to identify region of interest (ROI), then processing only the extracted ROI at full resolution. This segmentation of the image and processing steps reduces the computational burden while maintaining recognition accuracy for the important regions.
Solution Approach 2:
The patent implements local quality by applying different processing resolutions to different parts of the image. The full-resolution processing is applied only to the ROI extracted from the downsampled image, while the rest of the image is processed at lower resolution or not at all, optimizing the balance between accuracy and computational efficiency.
2Measurement precision
If high-resolution images are processed directly, then feature extraction quality is improved, but memory requirements and computational resources increase
Solution Approach 1:
The patent segments the image processing task to load only necessary portions into memory. By first identifying ROI in a downsampled image and then extracting only that region at full resolution, the system minimizes the amount of high-resolution image data that needs to be stored in memory during processing.
Solution Approach 2:
The patent extracts only the region of interest from the full-resolution image after identifying it through downsampled image processing. This extraction approach ensures that only the necessary portion of the high-resolution image is loaded into memory and processed, reducing overall memory requirements while maintaining feature extraction quality for the important regions.
3Reliability
If the entire high-resolution image is input to artificial intelligence models, then comprehensive object detection is achieved, but processing efficiency decreases
Solution Approach 1:
The patent segments the detection process into two phases: rapid ROI identification using a downsampled image, followed by detailed object detection only within the extracted ROI at full resolution. This segmentation maintains detection completeness for important objects while significantly reducing processing time compared to analyzing the entire high-resolution image.
Solution Approach 2:
The patent performs preliminary processing by downsampling the image and identifying ROI before conducting the main object detection task. This preliminary action of locating regions of interest in advance allows the subsequent detection process to focus only on relevant areas, improving overall processing efficiency while maintaining detection reliability.
Data Source
AI summary
An electronic apparatus may include a processor configured to: obtain an image captured by a camera; obtain a downscaled image by downscaling the captured image, wherein the downscaled image has an image that is less than a critical resolution, identify a region of interest included in the downscaled image by inputting the downscaled image into a first artificial intelligence model, the first artificial intelligence model being trained to identify a region of interest in an image; extract, from the captured image, an object image in the captured image corresponding to the identified region of interest, and obtain information on an object region included in the captured image by inputting the extracted object image into a second artificial intelligence model, the second artificial intelligence model being configured to identify an object region in an input image.