Image Processor Depth-Based Candidate Area Extraction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Processing shot images in environments with numerous objects is inefficient due to high processing loads caused by the need for high temporal and spatial resolution, leading to delayed response to subject motion.
Innovation Solution
An information processor with a candidate area extraction section, detailed information acquisition section, and output information generation section that uses template matching on depth images to efficiently identify and track target objects by extracting candidate areas and adjusting processing detail based on object distance.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If template matching is performed on the entire shot image to ensure high detection accuracy, then measurement precision is improved, but processing time increases and productivity decreases
Solution Approach 1:
The patent divides the shot image into multiple regions based on depth information, performing template matching only in candidate regions where targets are likely to exist. This segmentation approach maintains detection accuracy by focusing on relevant areas while significantly reducing the overall processing time compared to analyzing the entire image.
Solution Approach 2:
The patent applies different processing strategies to different regions of the image based on depth data. Candidate regions identified through depth-based filtering receive detailed template matching processing, while other regions are processed more quickly or skipped, creating local quality variations in processing intensity that optimize both accuracy and speed.
2Measurement precision
If high temporal and spatial resolution is used for target detection, then measurement precision is improved, but processing load increases
Solution Approach 1:
The patent performs preliminary filtering using depth information before conducting detailed template matching. By pre-identifying candidate regions where targets are likely to be located based on depth data, the system reduces the amount of high-resolution processing needed, thereby lowering the overall processing load while maintaining detection accuracy in the relevant areas.
Solution Approach 2:
The patent extracts and processes only the necessary portions of the image data - specifically, candidate regions identified through depth-based filtering. By taking out and processing only these relevant portions rather than the entire high-resolution image, the system maintains measurement precision where needed while significantly reducing the total processing load.
3Measurement precision
If detailed analysis is performed on the entire shot image, then measurement precision is improved, but response time to subject motion deteriorates
Solution Approach 1:
The patent segments the image analysis process into two stages: first, rapid identification of candidate regions using depth information, and second, detailed analysis only within those candidate regions. This segmentation enables the system to maintain high measurement precision in the regions that matter while dramatically reducing the time required compared to analyzing the entire image in detail.
Solution Approach 2:
The patent applies partial action by performing detailed analysis only in candidate regions rather than the entire image. This partial processing approach provides sufficient accuracy for the regions containing potential targets while avoiding the time cost of analyzing all image areas, thus improving response time to subject motion.
Data Source
AI summary
An image storage section 48 stores shot image data with a plurality of resolutions transmitted from an imaging device. Depth images 152 with a plurality of resolutions are generated using stereo images with a plurality of resolution levels from the shot image data (S10). Next, template matching is performed using a reference template image 154 that represents a desired shape and size, thus extracting a candidate area for a target picture having the shape and size for each distance range associated with one of the resolutions (S12). A more detailed analysis is performed on the extracted candidate areas using the shot image stored in the image storage section 48 (S14). In some cases, a further image analysis is performed based on the analysis result using a shot image with a higher resolution level (S16a and S16b).