Image Processing Apparatus Reducing VRAM Usage via Object Region Extraction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
When the number of captured images is large and their resolution is high, loading them onto graphics memory (VRAM) simultaneously leads to very high memory usage, potentially limiting VRAM usage or making it difficult to load all images due to capacity constraints.
Innovation Solution
The proposed image processing technique involves obtaining data from multiple captured images and camera parameters, generating object images by extracting image regions corresponding to the object from each captured image, and obtaining spatial information based on these object images and their corresponding camera parameters, thereby reducing memory usage in the VRAM.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If a plurality of captured images is loaded onto VRAM at the same time, then learning of spatial information can be performed, but memory usage becomes very high
Solution Approach 1:
The captured images are segmented into multiple patches, where only the patches containing the object of interest are extracted and processed. This segmentation approach reduces the total data volume loaded onto VRAM while preserving the essential information needed for spatial learning, thereby resolving the contradiction between learning accuracy and memory usage.
Solution Approach 2:
The patent extracts only the necessary object regions from the captured images, removing unnecessary background and other elements. This extraction process significantly reduces the memory footprint of the input data while maintaining the quality of spatial information learning, directly addressing the memory usage problem.
2Manufacturing precision
If high-resolution captured images are loaded onto VRAM, then processing quality is improved, but VRAM capacity is exceeded
Solution Approach 1:
High-resolution captured images are divided into smaller patches, allowing selective processing of only those patches containing the object. This segmentation enables maintaining high processing quality for relevant regions while significantly reducing the total VRAM capacity required, as only essential patches are loaded.
Solution Approach 2:
The patent applies different processing quality levels to different regions: high-quality processing is applied to patches containing the object of interest, while patches without the object are either discarded or processed at lower quality. This local quality approach optimizes the balance between processing quality and VRAM capacity utilization.
Data Source
AI summary
The image processing apparatus obtains data of a plurality of captured images obtained by capturing an object from a plurality of directions and image capturing camera parameters, which are camera parameters corresponding to each of the plurality of captured images, generates a plurality of object images by extracting an image region corresponding to an image of the object from each of the plurality of captured images, generates object image camera parameters, which are camera parameters corresponding to each of the plurality of object images, based on the image capturing camera parameters and information indicating a position of an image region corresponding to the image of the object in the plurality of captured images, and obtains spatial information representing a space in which the object exists based on the plurality of object images and the object image camera parameters corresponding to each of the plurality of object images.