Binocular Depth Estimation Mapping for Feature-Poor Image Regions

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Existing image depth estimation methods, such as binocular ranging, can only determine depth for image portions with prominent feature points, failing to estimate depth for areas lacking distinct features.

Innovation Solution

An image depth estimation method that calculates relative depth values for multiple pixels in a binocular image, identifies a target region with consistent bidirectional disparity, determines absolute depth values within this region, constructs a mapping relationship between relative and absolute depth values, and applies this mapping to estimate depth across the entire image.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If binocular ranging method is used to estimate depth, then depth can be determined for image portions with prominent feature points, but depth cannot be determined for portions with non-prominent feature points

Engineering Contradiction:
Improvedepth estimation accuracyVSAvoidapplicability to various image types
Core Design Contradiction:
Measurement precisionVSAdaptability or versatility

Solution Approach 1:

The patent segments the image into different regions based on feature point prominence. Prominent feature point regions are processed using traditional binocular ranging methods, while non-prominent feature point regions are processed using a different approach that leverages spatial relationships and depth maps from prominent regions to infer depths in these areas, thereby achieving comprehensive depth estimation across the entire image.

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

The patent introduces an intermediary mechanism that uses the depth information and spatial relationships derived from prominent feature point regions as mediators to infer and estimate depths in non-prominent feature point regions. This intermediary approach allows the system to bridge the gap between regions with and without prominent features, enabling unified depth estimation across diverse image types.

Inventive Principle:
Principle #24Intermediary (Mediator)

2Productivity

If depth estimation is performed only on image portions with prominent feature points, then processing complexity is reduced, but complete depth estimation for the entire image is not achieved

Engineering Contradiction:
Improvedepth estimation coverageVSAvoidprocessing complexity
Core Design Contradiction:
ProductivityVSDevice complexity

Solution Approach 1:

The patent performs preliminary processing to identify and segment prominent feature point regions first, establishing depth information and spatial relationships in these key areas. This preliminary action creates a foundation that simplifies the subsequent inference process for non-prominent regions, allowing the system to achieve complete image coverage without proportionally increasing overall processing complexity.

Inventive Principle:
Principle #10Preliminary action

Solution Approach 2:

The patent applies different processing strategies to different local regions of the image based on their feature point characteristics. Prominent feature regions receive detailed binocular ranging processing, while non-prominent regions receive simplified inference-based processing. This local quality differentiation allows the system to maximize depth estimation coverage while managing processing complexity through region-specific approaches.

Inventive Principle:
Principle #3Local quality

Data Source

PatentEP4589531A1Image depth estimation method and apparatus, electronic device and storage medium
Publication Date: 2025.07.23 BEIJING ZITIAO NETWORK TECH CO LTD
  • EP4589531A1 patent drawingFigure 1
  • EP4589531A1 patent drawingFigure 2
  • EP4589531A1 patent drawingFigure 3~4

AI summary

Embodiments of the present disclosure disclose an image depth estimation method and apparatus, an electronic device, and a storage medium. The method includes: obtaining relative depth values of a plurality of pixels in a binocular image; obtaining a bidirectional disparity of a same scene point in the binocular image, and determining a target region in which the bidirectional disparity is consistent; determining an absolute depth value of a pixel in the target region based on a bidirectional disparity of the pixel in the target region; constructing a mapping relationship between a relative depth value and an absolute depth value based on the absolute depth value of the pixel in the target region and a relative depth value of the pixel in the binocular image; and determining, based on the mapping relationship and a relative depth value of a pixel to be determined in the binocular image, an absolute depth value of the pixel to be determined in the binocular image.