Depth Map Reprojection for Temporally Consistent Stereo Estimation

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Conventional mixed-reality systems face challenges in generating depth maps with high computational expense, latency, imprecise depth borders, and temporal inconsistencies, leading to parallax errors and artifacts in parallax-corrected images.

Innovation Solution

The system employs a method for temporally consistent depth map generation by obtaining a first stereo pair of images, generating a first depth map, and using a reprojected depth map to align with a second stereo pair of images, thereby improving temporal consistency and overall depth map quality.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If conventional stereo matching algorithms are used to generate depth maps, then depth information can be obtained, but computational expense increases and latency occurs

Engineering Contradiction:
Improvedepth map accuracyVSAvoidlatency
Core Design Contradiction:
Measurement precisionVSLoss of time

Solution Approach 1:

The system performs preliminary actions by reprojecting the previous depth map to the current stereo pair's viewpoint before generating the new depth map. This preliminary reprojection creates a temporal reference that guides the current depth estimation, reducing the computational search space and accelerating the matching process while maintaining accuracy.

Inventive Principle:
Principle #10Preliminary action

Solution Approach 2:

The system maintains continuity by incorporating the previous depth map into the current frame's depth estimation process. Instead of independently processing each frame, the depth information flows continuously across frames through reprojection and fusion, eliminating redundant computations and reducing latency while preserving measurement precision.

Inventive Principle:
Principle #20Continuity of useful action

2Measurement precision

If conventional stereo matching algorithms are used to generate depth maps, then depth information can be obtained, but temporal inconsistencies occur leading to depth flickers

Engineering Contradiction:
Improvedepth map consistencyVSAvoidtemporal consistency
Core Design Contradiction:
Measurement precisionVSStability of the object's composition

Solution Approach 1:

The system implements feedback by using the reprojected previous depth map as a reference guide for the current depth estimation. The algorithm compares current stereo disparities with the reprojected historical depth information and uses this feedback to constrain and stabilize the depth estimation, preventing temporal flickers while maintaining precision.

Inventive Principle:
Principle #23Feedback

Solution Approach 2:

By reprojecting the previous depth map to the current viewpoint before generating the new depth map, the system establishes a temporal reference in advance. This preliminary action creates a stable baseline that guides the current estimation and ensures temporal consistency, preventing flickers caused by independent frame processing.

Inventive Principle:
Principle #10Preliminary action

3Measurement precision

If depth maps are generated for each stereo pair, then depth information is available, but computational expense increases

Engineering Contradiction:
Improvedepth map qualityVSAvoidcomputational expense
Core Design Contradiction:
Measurement precisionVSUse of energy by moving object

Solution Approach 1:

The system merges the current stereo matching results with the reprojected previous depth map to generate the final depth map. This combination allows the algorithm to leverage historical depth information, reducing the computational effort required for current frame processing while maintaining high depth map quality through the fusion of multiple sources.

Inventive Principle:
Principle #5Merging (Combining)

Solution Approach 2:

The reprojection of the previous depth map serves as a preliminary computation that prepares a reference solution in advance. This preliminary action reduces the computational burden of the current depth estimation by providing a starting point or constraint, thereby lowering overall computational expense while preserving depth map quality.

Inventive Principle:
Principle #10Preliminary action

Data Source

PatentEP4150578B1Systems and methods for temporally consistent depth map generation
Publication Date: 2025.09.24 MICROSOFT TECHNOLOGY LICENSING LLC
  • EP4150578B1 patent drawingFigure 1
  • EP4150578B1 patent drawingFigure 2
  • EP4150578B1 patent drawingFigure 3

AI summary

Systems and methods are provided for performing temporally consistent depth map generation by implementing acts of obtaining a first stereo pair of images of a scene associated with a first timepoint and a first pose, generating a first depth map of the scene based on the first stereo pair of images, obtaining a second stereo pair of images of the scene associated with at a second timepoint and a second pose, generating a reprojected first depth map by reprojecting the first depth map to align the first depth map with the second stereo pair of images, and generating a second depth map that corresponds to the second stereo pair of images using the reprojected first depth map.