Image Resolution Enhancement via Object Importance Segmentation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional image resolution enhancement techniques using machine learning often fail to achieve sufficient resolution for important objects in a timely manner, leading to extended processing times as target resolution increases, necessitating a compromise between resolution and processing speed.
Innovation Solution
An image processing apparatus that determines conversion parameters for enhancing image resolution based on object importance, using a dataset generated from object-specific information, allowing for adjustable resolution enhancement levels and efficient processing by prioritizing high-resolution enhancement for critical objects while reducing processing time.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Manufacturing precision
If the target resolution is set high to obtain sufficient image resolution, then the manufacturing precision (image resolution quality) is improved, but the loss of time (processing time) increases
Solution Approach 1:
The patent applies local quality by differentiating resolution enhancement processing based on object importance. Important objects receive high-resolution enhancement with full processing time, while less important objects use lower resolution or reduced processing. This allows the system to achieve high manufacturing precision for critical objects without incurring the full time penalty across all objects, thus resolving the contradiction between image resolution quality and processing time.
Solution Approach 2:
The patent segments the image processing task by identifying and separating important objects from other regions. The determination unit divides the image into multiple regions based on object importance, allowing different processing strategies to be applied to different segments. This segmentation enables high-resolution enhancement to be focused only where necessary, reducing overall processing time while maintaining high resolution quality for important objects.
2Loss of time
If the target resolution is set low to reduce processing time, then the loss of time is reduced, but the manufacturing precision (image resolution quality) deteriorates
Solution Approach 1:
By applying local quality differentiation, the patent ensures that important objects always receive high-resolution processing regardless of overall time constraints. The system dynamically adjusts processing quality based on local object importance rather than applying a uniform low-resolution setting, thus preventing deterioration of image resolution quality for critical regions even when processing time is reduced.
Solution Approach 2:
The segmentation approach allows the system to process only important objects at high resolution while using faster, lower-resolution methods for other regions. This selective processing maintains high manufacturing precision for important objects while significantly reducing the total processing time by avoiding full high-resolution processing across the entire image.
3Measurement precision
If learning is performed for each class to improve recognition accuracy, then the measurement precision (recognition accuracy) is improved, but the device complexity increases
Solution Approach 1:
The patent applies partial action by performing detailed learning and classification only for important objects rather than all objects in the image. The determination unit identifies which objects require high-precision recognition, and learning is concentrated on these partial cases. This reduces the overall device complexity and computational burden while maintaining high measurement precision for the most critical recognition tasks.
Data Source
AI summary
The image processing apparatus has: an image acquisition unit configured to acquire a captured image of an image capturing area in which an object is located; a determination unit configured to determine parameters to be used for image processing to improve resolution of an image of the object, by learning using a dataset of images of the object, wherein the dataset is generated based on object information indicating a degree of importance of the object; and a processing unit configured to perform the image processing to improve the resolution of an image of the object included in the acquired captured image, using the parameters determined by the determination unit.