Mask Super-Resolution for Precise Object Extraction Edges
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing neural network models for image processing suffer from low precision, leading to issues such as sawteeth at the edges of extracted target objects in large images.
Innovation Solution
An image processing method involving downsampling an original image, extracting a target object region to obtain a mask, using a mask super-resolution model for super-resolution processing, and fusing the processed mask with the original image to improve resolution and avoid edge artifacts.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Extent of automation
If a neural network model is used for target detection and extraction, then the processing can be automated and widely applied to various images, but the processing precision is low and sawteeth occur at the edge of processed images
Solution Approach 1:
The processing is segmented into two stages: first processing a downsampled version of the image to obtain initial target region, then using super-resolution technology to refine the mask at higher resolution. This segmentation allows automated processing while improving edge precision by handling different resolution levels separately.
Solution Approach 2:
The target region is preliminarily identified in the downsampled image before super-resolution processing. This preliminary action allows the system to focus computational resources on refining only the relevant target regions at high resolution, improving both automation efficiency and edge quality.
2Manufacturing precision
If the original high-resolution image is processed directly, then the processing precision would be high, but the computational complexity and processing time would increase significantly
Solution Approach 1:
The image processing is segmented by resolution level: downsampled image processing for target identification, and super-resolution processing only for the target mask refinement. This avoids the computational burden of processing the entire high-resolution image while maintaining precision where it matters most.
Solution Approach 2:
High processing quality is applied locally only to the target object mask through super-resolution processing, while the rest of the image is handled at lower resolution. This ensures high precision for the target extraction while maintaining overall processing efficiency.
3Manufacturing precision
If super-resolution processing is applied to the target mask, then the resolution and fineness of the mask is improved, but the processing time and computational resources increase
Solution Approach 1:
The target region is preliminarily identified in the downsampled image before applying super-resolution processing. This preliminary identification ensures that super-resolution computation is applied only to the necessary target mask regions, reducing overall processing time while maintaining high mask resolution.
Solution Approach 2:
Super-resolution processing is applied partially only to the target mask rather than the entire image. This partial action provides sufficient fineness for the target extraction without the excessive computational cost of full-image super-resolution processing.
Data Source
AI summary
The present disclosure provides an image processing method, including: down-sampling an original image according to a preset resolution to generate a down-sampled image; extracting a target object region from the downsampled image to obtain a first target object mask; inputting the downsampled image and the first target object mask into a mask super-resolution model obtained in advance by training; performing super-resolution processing on the first target object mask by using the mask super-resolution model to obtain a second target object mask; and fusing the second target object mask with the original image to obtain a target object image. The present disclosure further provides an electronic device and a non-transitory computer readable medium.