Image Segmentation and Style Transfer for Mobile Devices
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing image processing methods face challenges in accurately segmenting objects with abnormal postures, distinguishing object categories, handling small object proportions, and achieving real-time style transfer on mobile devices due to limitations in processing speed and edge fusion effects.
Innovation Solution
An image processing method that segments images to identify target regions, predicts and corrects rotation angles, uses neural networks for object segmentation, and employs a generative adversarial network for style transfer, optimizing the network structure for mobile devices to enhance processing efficiency and accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If existing image processing methods are used for object segmentation, then processing can be performed on standard images, but segmentation accuracy deteriorates for objects with abnormal postures and small proportions
Solution Approach 1:
The image processing method segments the image into multiple regions including foreground objects, background, and edge regions. By separating edge regions and processing them with specialized fusion techniques, the method improves segmentation accuracy for objects with abnormal postures and small proportions that would otherwise be missed or misclassified by standard segmentation algorithms.
Solution Approach 2:
Different processing strategies are applied to different regions of the image. Edge regions are handled with specialized fusion techniques that preserve fine details and boundaries, while central regions use standard segmentation. This local differentiation allows the system to maintain high accuracy for small and oddly-shaped objects without compromising overall processing efficiency.
2Manufacturing precision
If complex neural networks are used for style transfer, then style transfer quality improves, but processing speed deteriorates on mobile devices
Solution Approach 1:
The network processes only the segmented target regions rather than entire images. By limiting style transfer computation to relevant foreground objects and excluding background areas, the method maintains high style transfer quality while reducing the computational load and processing time required on mobile devices.
Solution Approach 2:
The method extracts and processes only the essential style transfer components from complex images. By identifying and isolating key visual features that need style transfer application, the system achieves high-quality results with reduced computational requirements, enabling real-time processing on mobile platforms.
3Measurement precision
If rotation angle correction is not performed, then processing is simpler and faster, but segmentation accuracy deteriorates for rotated objects
Solution Approach 1:
The system performs preliminary detection of object rotation angles before conducting the main segmentation process. By predicting rotation angles in advance and applying corrective transformations, the method ensures that subsequent segmentation operations work with properly oriented objects, significantly improving segmentation accuracy for rotated objects without adding excessive complexity to the overall system.
Data Source
AI summary
The disclosure relates to a communication method and system for converging a 5th-Generation (5G) communication system for supporting higher data rates beyond a 4th-Generation (4G) system with a technology for IoT. The disclosure may be applied to intelligent services based on the 5G communication technology and the IoT-related technology, such as smart home, smart building, smart city, smart car, connected car, health care, digital education, smart retail, security and safety services. An image processing method and apparatus, electronic device and computer readable storage medium, which belong to image processing field are provided. The method and apparatus, electronic device and computer readable storage medium include segmenting an image to be processed to obtain a target region in the image to be processed, and performing style transfer on the target region. The solution provided may effectively improve effects of image processing, and better meet requirements of practical application.