Hierarchical Neural Displacement Map Generation

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Existing methods for generating disparity maps from stereo images are computationally costly and complex, with high implementation complexity and high calculation costs, limiting their effectiveness in real-world applications such as autonomous vehicles and augmented reality.

Innovation Solution

A method and apparatus using a neural network-based feature extractor to generate a displacement map by processing input datasets, employing a hierarchical feature map hierarchy with refinement pairs to produce high-quality disparity maps efficiently, while reducing computational costs through strategic upsampling and warping operations.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If neural networks are used for generating disparity maps from stereo images, then measurement precision is improved, but device complexity and calculation costs increase

Engineering Contradiction:
Improvedisparity map accuracyVSAvoidimplementation complexity
Core Design Contradiction:
Measurement precisionVSDevice complexity

Solution Approach 1:

The patent segments the disparity map generation process into multiple hierarchical levels, processing images at different resolutions (e.g., 1/4, 1/8, 1/16 of original size). This segmentation allows neural networks to operate on smaller, more manageable data sets at each level, reducing computational complexity while maintaining accuracy through progressive refinement from coarse to fine details.

Inventive Principle:
Principle #1Segmentation

2Measurement precision

If high-resolution disparity maps are generated directly, then measurement precision is improved, but calculation costs and processing time increase

Engineering Contradiction:
Improvedisparity map resolutionVSAvoidprocessing time
Core Design Contradiction:
Measurement precisionVSLoss of time

Solution Approach 1:

The patent applies preliminary action by first generating low-resolution disparity maps at coarser hierarchical levels before processing higher resolutions. The coarse-level disparity information serves as a preliminary result that guides and constrains subsequent fine-level processing, reducing the search space and computational burden at each refinement stage while ensuring final high-resolution accuracy.

Inventive Principle:
Principle #10Preliminary action

3Measurement precision

If multiple hierarchical levels of feature maps are processed, then measurement precision is improved, but device complexity increases

Engineering Contradiction:
Improvedisparity estimation accuracyVSAvoidsystem complexity
Core Design Contradiction:
Measurement precisionVSDevice complexity

Solution Approach 1:

The patent merges results from multiple hierarchical levels through a unified processing framework. Disparity information from different resolution levels is combined and refined systematically, where coarser levels provide global structure and finer levels add local detail. This merging approach achieves high precision while managing complexity through structured integration rather than independent processing of each level.

Inventive Principle:
Principle #5Merging (Combining)

Data Source

PatentEP3803791B1Generating a displacement map of an input dataset pair of image or audio data
Publication Date: 2023.11.22 AIMOTIVE KFT
  • EP3803791B1 patent drawingFigure 1
  • EP3803791B1 patent drawingFigure 2
  • EP3803791B1 patent drawingFigure 3

AI summary

The invention is a method and an apparatus for generating a displacement map of a first input dataset and a second input dataset of an input dataset pair. The apparatus comprises - a neural network based feature extractor (25) for processing the first input dataset and the second input dataset so as to generate a feature map hierarchy (50) comprising a base pair of feature maps (20e, 30e) and a refinement pair of feature maps (20b, 30b, 20c, 30c, 20d, 30d), - a displacement unit comprising a first comparator unit for obtaining an initial displacement map using the base pair of feature maps (20e, 30e); and - a displacement refinement unit for obtain an updated displacement map for the refinement pair of feature maps (20d, 30d).