Object Region Selection Using Likelihood Maps and Region Tensors
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing object detection systems in image processing face challenges in accurately identifying the intended subject due to errors in estimating the likelihood and vector positioning, especially when user-designated coordinates deviate from the actual subject, leading to misidentification.
Innovation Solution
An image processing apparatus that includes an obtaining unit for captured images, a likelihood map unit, a region tensor unit, and a region determining unit to accurately identify the intended subject by integrating and correcting user-designated coordinates using a likelihood map and region tensor, thereby enhancing the accuracy of object region detection.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of operation
If user-designated coordinates are used directly for object identification, then the operation is simple and fast, but the identification accuracy deteriorates when coordinates are deviated from the actual subject
Solution Approach 1:
The patent introduces an intermediary processing step between user coordinate input and subject identification. The system uses the designated coordinates as a starting point to search for and identify the actual object center, rather than directly using the coordinates. This intermediary process acts as a mediator that translates imprecise user input into accurate subject identification.
Solution Approach 2:
The system implements feedback by using the designated coordinates to search for the actual object center, then comparing and adjusting based on the detected object position. The likelihood map and region tensor provide feedback information that guides the correction of initial coordinate estimates, creating a closed-loop system that improves accuracy through iterative refinement.
2Measurement precision
If additional processing steps are added to correct coordinate deviations, then the identification accuracy is improved, but the device complexity increases
Solution Approach 1:
The patent performs preliminary actions by pre-calculating and storing likelihood maps and region tensors before actual object identification. These pre-computed data structures contain information about object positions and characteristics, allowing the system to quickly correct coordinate deviations without adding complex real-time processing. The preliminary preparation of these data structures simplifies the main identification process.
Solution Approach 2:
The system replaces complex mechanical or algorithmic coordinate correction mechanisms with data-driven approaches using likelihood maps and region tensors. Instead of implementing complex iterative optimization algorithms, the system substitutes these with pre-computed probability distributions and region information that can be efficiently queried and processed, reducing computational complexity while maintaining accuracy.
3Manufacturing precision
If the system uses likelihood maps and region tensors to correct coordinates, then the object region detection accuracy is improved, but the processing time and computational resources increase
Solution Approach 1:
The likelihood maps and region tensors are computed in advance and stored for quick retrieval during object identification. This preliminary computation approach allows the system to have accurate object information ready before actual detection is needed, significantly reducing the processing time during the main identification operation. The heavy computational work is performed beforehand when the system has more available resources.
Data Source
AI summary
An image processing apparatus comprises: an obtaining unit configured to obtain a captured image; a likelihood obtaining unit configured to obtain a likelihood map indicating a likelihood of existence of an object at each position of the captured image; a region obtaining unit configured to obtain, at each position of the captured image, a region tensor indicating a position and a size of an object with respect to each position; an accepting unit configured to accept first position coordinates with respect to the captured image; and a region determining unit configured to determine an object region corresponding to the first position coordinates, based on the region tensor and likelihoods of the likelihood map corresponding to two or more region candidates each indicated by the region tensor based on the first position coordinate.