Image Orientation Detection via Character Area Sampling
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for determining the orientation of images containing text on portable devices often result in undesirable rotation, as they rely solely on device orientation, which can lead to inaccurate alignment and detract from the user experience.
Innovation Solution
A method that samples character areas within an image to determine their orientation, using a character-level orientation classifier or regression model, allowing for more precise alignment of the image with the device pose without recognizing the characters, thereby reducing computational load and power consumption.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Extent of automation
If device orientation is used to determine image orientation, then the image rotation follows device rotation automatically, but the image information may not need to be rotated simultaneously leading to inaccurate alignment
Solution Approach 1:
The patent segments the image into multiple character areas and samples orientations from different regions independently. By dividing the image into multiple zones and analyzing character orientations in each zone, the system achieves more accurate overall image orientation determination compared to relying solely on device orientation sensors.
Solution Approach 2:
The patent replaces the mechanical sensor-based orientation detection (gravity sensor) with an image-processing-based orientation detection system. Instead of using physical sensors to determine orientation, the system analyzes the actual image content (character areas) to determine the correct orientation, substituting mechanical detection with computational analysis.
2Measurement precision
If character recognition is performed to determine orientation, then precise orientation can be achieved, but computational load and power consumption increase
Solution Approach 1:
The patent extracts only the essential orientation information from character areas without performing full character recognition. By sampling character areas and determining their orientation directly without identifying what the characters actually say, the system obtains precise orientation data while avoiding the high computational cost of complete character recognition.
Solution Approach 2:
The patent performs partial character area sampling rather than analyzing the entire image in detail. By selectively sampling representative character areas and determining orientation from these samples, the system achieves sufficient accuracy with reduced computational effort compared to comprehensive image analysis.
Data Source
AI summary
Embodiments of the present disclosure relate to a new approach for image orientation detection. In the computer-implemented method, at least one character area in an image is sampled. The orientation of the sampled character area is determined, and the orientation of the image is determined based on the determined orientation of the sampled character area.