Image Direction Judging Device Using Object Similarity
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional image direction determining devices face challenges in accurately determining the up-and-down direction of images without incorporating a slant detector, leading to increased costs and inaccurate determinations due to reliance on face detection or character recognition, which are prone to errors from facial expressions, illumination, and low detection precision.
Innovation Solution
An image direction determining device that detects objects with fixed shapes or arrangements, calculates similarity degrees between detected objects and pre-stored registration objects, and determines the up-and-down direction based on the angles and similarity degrees, allowing for accurate determination without the need for a slant detector.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If a slant detector is incorporated into the image acquiring device to determine the up-and-down direction, then the determination accuracy is improved, but the production cost increases
Solution Approach 1:
The patent replaces the mechanical slant detector with a computational image processing system. The up-and-down direction determination is achieved by detecting objects with fixed shapes or arrangements in the image, calculating their angles relative to the image frame, and determining the up-and-down direction based on these angle calculations, thereby eliminating the need for additional hardware sensors.
Solution Approach 2:
The patent uses pre-stored registration objects that represent the expected appearance and orientation of objects with fixed shapes or arrangements. By comparing detected objects against these registered templates, the system can determine image orientation without physical sensors, using digital pattern matching instead of mechanical detection.
2Ease of manufacture
If face detection or character recognition is used to determine the up-and-down direction, then the determination can be performed without a slant detector, but the accuracy decreases due to errors from facial expressions, illumination, and low detection precision
Solution Approach 1:
The patent focuses on detecting specific local features - objects with fixed shapes or arrangements - rather than relying on overall image content like face detection. By targeting objects with known, invariant geometric properties, the system achieves more reliable angle measurements that are not affected by variations in illumination, expression, or recognition accuracy.
Solution Approach 2:
The patent changes the detection parameter from semantic recognition (face detection, character recognition) to geometric parameter extraction (angle calculation of objects with fixed shapes). This parameter transformation from high-level recognition to low-level geometric measurement provides more stable and accurate results for orientation determination.
Data Source
AI summary
Up and down directions of an image are to be precisely judged without a special equipment installed in an image pickup device. An object candidate detecting means detects object candidates from an input image and their angles in the input image. A similarity calculation means calculate the degree of similarity between each detected object candidate and each object stored in advance. An input image angle calculating means judges up and down directions of the input image based on the calculated similarity of each object candidate and the angle of the input image. The input image angle calculating means carries out weighting for the angle of each object candidate in the input image based on its similarity and calculates slant angles with respect to the up and down directions of the input image by using the weighted angles.