Image Interpretation via Dynamic Priority Queue Optimization

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Existing image interpretation methods face computational challenges in efficiently localizing object parts due to the need to model spatial relations, which introduces dependencies among model variables, leading to exponential search complexity as the number of parts increases.

Innovation Solution

An iterative algorithm that uses a factorizable upper bound and dynamic priority queue ordering to efficiently optimize the objective function, allowing for the efficient localization of object parts by evaluating candidate configurations in descending order of their upper bound scores.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If spatial relations are modeled among object parts, then localization performance is improved, but computational complexity increases exponentially

Engineering Contradiction:
Improvelocalization performanceVSAvoidcomputational complexity
Core Design Contradiction:
Measurement precisionVSDevice complexity

Solution Approach 1:

The patent segments the joint optimization problem into iterative local optimizations for each part. Instead of optimizing all K parts simultaneously (which would require evaluating N^K configurations), the algorithm optimizes one part at a time while holding others fixed, reducing the complexity from exponential N^K to linear K×N evaluations.

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

The patent introduces dynamic iterative optimization where the configuration of parts is updated sequentially in multiple passes. The algorithm dynamically adjusts part locations by evaluating candidate positions and updating the configuration iteratively, allowing the system to converge to an optimal solution without exhaustive search.

Inventive Principle:
Principle #15Dynamics

2Reliability

If the number of object parts in the model is increased, then detection accuracy is improved, but search time becomes intractable

Engineering Contradiction:
Improvedetection accuracyVSAvoidsearch time
Core Design Contradiction:
ReliabilityVSLoss of time

Solution Approach 1:

The patent divides the search problem into independent sub-problems for each part. By segmenting the joint configuration search into individual part optimizations, the algorithm can handle models with many parts (K parts with N candidates each) efficiently, requiring only K×N evaluations rather than N^K exhaustive searches.

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

The patent performs partial optimization by iteratively refining part configurations rather than requiring complete simultaneous optimization. The algorithm performs multiple passes through the parts, updating configurations incrementally, which allows accurate detection with significantly reduced computational effort compared to exhaustive search.

Inventive Principle:
Principle #16Partial or excessive action

3Productivity

If independent part modeling is used, then computational efficiency is improved, but localization accuracy deteriorates

Engineering Contradiction:
Improvecomputational efficiencyVSAvoidlocalization accuracy
Core Design Contradiction:
ProductivityVSMeasurement precision

Solution Approach 1:

The patent introduces dynamic iterative updates that allow parts to influence each other's positions over multiple passes. While parts are optimized independently in each iteration, the iterative process allows spatial relationships to emerge as each part's position is refined based on the current configuration of others, achieving both efficiency and accuracy.

Inventive Principle:
Principle #15Dynamics

Solution Approach 2:

The patent maintains continuous refinement of part configurations through multiple iterative passes. Instead of a single independent evaluation, the algorithm continuously updates part positions across iterations, allowing the system to converge to accurate localized configurations while maintaining computational efficiency through independent part optimization in each pass.

Inventive Principle:
Principle #20Continuity of useful action

Data Source

PatentEP2390822B1System and method for efficient interpretation of images in terms of objects and their parts
Publication Date: 2019.04.24 PALO ALTO RESEARCH CENTER INC
  • EP2390822B1 patent drawingFigure 1
  • EP2390822B1 patent drawingFigure 2
  • EP2390822B1 patent drawingFigure 3

AI summary

The present application is a method and system of interpreting an image by finding a configuration of multiple variables which optimizes an objective function with a factorizable upper bound, by applying an iterative algorithm that relies on efficient dynamic ordering of candidate configurations, in a priority queue, in a descending order of an upper bound score. As an example, consider a constellation model for an object. It specifies the appearance models f or individual parts of objects, as well as spatial relations among these parts. These are combined into a single function whose value represents the likeness of the object in an image. To find the configuration in which the object is present in the image, we maximize this function over all candidate configurations. The purpose of the iterative algorithm mentioned above is to find such optimal configurations efficiently.