Object Region Recognition Using Edge-Semantic Image Fusion
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional video editing systems lack the ability to efficiently detect and recognize object regions in image frames, limiting their applicability in editing tasks such as replacing trademarks or logos, as they primarily rely on text recognition models that fail to capture graphical elements.
Innovation Solution
A method involving edge detection, semantic segmentation, and integration of edge and semantic region features using a residual network model to accurately identify object regions, enhancing the recognition of graphical elements with lower computational complexity.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If conventional video editing techniques are used, then video editing can be performed, but the ability to detect and recognize image regions in video frames is lacking
Solution Approach 1:
The patent combines edge detection technology with semantic segmentation technology to create an integrated object region recognition system. The edge detection module extracts edge information from video frames, while the semantic segmentation module processes semantic information, and both are fused to achieve accurate object region recognition, thereby enhancing video editing capabilities with region detection functionality.
Solution Approach 2:
The patent introduces an object region recognition module as an intermediary between conventional video editing techniques and image region detection. This module acts as a bridge that processes video frames to identify and locate specific objects, providing the missing detection capability while maintaining compatibility with existing video editing workflows.
2Measurement precision
If edge detection and semantic segmentation are integrated to recognize object regions, then recognition accuracy is improved, but system complexity increases
Solution Approach 1:
The patent divides the object region recognition system into distinct functional modules: an edge detection module that extracts edge information, a semantic segmentation module that processes semantic features, and a fusion module that integrates both types of information. This modular segmentation allows each component to specialize in specific tasks, improving overall recognition accuracy while managing system complexity through organized functional separation.
Solution Approach 2:
The patent designs the object region recognition system to perform multiple functions simultaneously: edge detection, semantic segmentation, and information fusion. By creating a multi-functional system that can handle different types of image analysis tasks within a unified framework, the patent achieves high recognition accuracy without proportionally increasing system complexity, as the same infrastructure supports multiple operations.
Data Source
AI summary
Embodiments of the present disclosure provide a method, an electronic device, and a computer program product for recognizing object regions in an image. The method includes detecting edges in an image based on pixels in the image. The method further includes segmenting the image into a plurality of semantic regions based on semantics of objects in the image. The method further includes recognizing, based on the edges and the plurality of semantic regions, object regions covering the objects. In this way, the positions of objects including graphic elements in the image and the tags of the objects can be automatically recognized, thus improving the efficiency of subsequent processing of the image.