Semantic Parsing of Objects in Video Using Multi-Resolution Analysis
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current methods for body parsing in video surveillance only estimate the localization of object parts without efficiently estimating semantic attributes associated with these parts, leading to inefficiencies and inaccuracies in identifying objects.
Innovation Solution
A method that produces multiple versions of an image at different resolutions, computes appearance scores, determines attribute configurations, and displays these configurations to simultaneously estimate both part locations and semantic attributes using appearance, resolution context, and geometric features.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Device complexity
If only syntactic object parsing is performed to estimate part locations, then the computational complexity is reduced, but the semantic attribute estimation accuracy deteriorates
Solution Approach 1:
The patent segments the image processing into multiple resolution levels (coarse to fine) and divides semantic attribute estimation into separate stages. Different attribute types (color, texture, shape) are estimated at appropriate resolution levels, reducing overall computational complexity while maintaining accuracy.
Solution Approach 2:
The patent introduces a resolution dimension by processing images at multiple scales. Semantic attributes are estimated across different resolution levels, allowing the system to capture both global context and local details without requiring full-resolution processing for all attribute types.
2Measurement precision
If multiple image versions at different resolutions are produced, then the semantic attribute estimation accuracy is improved, but the computational resources and time required increase
Solution Approach 1:
The patent performs preliminary processing by creating downsampled versions of images at lower resolutions before full-resolution analysis. This preliminary multi-resolution processing allows for efficient coarse-to-fine attribute estimation, reducing the computational burden on subsequent detailed analysis stages.
Solution Approach 2:
The patent applies partial processing by estimating only certain semantic attributes at reduced resolutions rather than processing all attributes at full resolution. This selective approach maintains accuracy for critical attributes while reducing overall computational requirements.
3Speed
If only part localization is performed without semantic attribute estimation, then the processing speed is maintained, but the object identification completeness deteriorates
Solution Approach 1:
The patent applies local quality by estimating different semantic attributes (color, texture, shape) at different resolution levels appropriate to each attribute type. This allows efficient processing speed for global attributes while maintaining identification completeness through detailed local attribute estimation where needed.
Data Source
AI summary
Methods, systems, and computer program products for parsing objects in a video are provided herein. A method includes producing a plurality of versions of an image of an object derived from a video input, wherein each version has a different resolution of said image of said object; computing an appearance score at each of a plurality of regions on the lowest resolution version of said plurality of versions of said image for at least one attribute for said object, wherein said appearance score denotes a probability of the at least one attribute appearing in the region; determining a configuration of the at least one attribute in the lowest resolution version based on at least the appearance score in each of the plurality of regions in the lowest resolution version; and displaying said configuration.