Depth Image Filtering with Scene-Adaptive Parameter Selection

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Existing depth image processing methods face challenges in achieving high-quality filtering due to the need for manual parameter setting adjustments based on scene types, leading to issues like fly pixel points, especially for users without professional experience.

Innovation Solution

Adaptive filtering is performed on depth images by determining scene types based on depth and confidence coefficients, allowing for selective filtering modes to improve image quality.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Manufacturing precision

If general filtering algorithms (bilateral filtering, anisotropic filtering, fixed threshold filtering) are used on depth images, then filtering process can be performed, but manual parameter setting is required for different scene types which is difficult for users without professional experience and causes fly pixel points

Engineering Contradiction:
Improvedepth image qualityVSAvoidparameter setting difficulty
Core Design Contradiction:
Manufacturing precisionVSEase of operation

Solution Approach 1:

The system automatically determines scene type based on depth image statistics (depth distribution, confidence coefficient distribution) and selects appropriate filtering parameters without user intervention. The algorithm self-adjusts parameters like filtering threshold and kernel size based on the detected scene characteristics, making the system easy to operate while maintaining high depth image quality

Inventive Principle:
Principle #25Self-service

Solution Approach 2:

The system dynamically changes filtering parameters based on detected scene type. Different scene types (close-range, mid-range, far-range, complex scenes) trigger different parameter configurations including filtering threshold, kernel size, and algorithm selection. This adaptive parameter adjustment eliminates the need for manual setting while optimizing depth image quality for each specific scenario

Inventive Principle:
Principle #35Parameter changes

2Manufacturing precision

If filtering parameters are manually adjusted for specific scenarios, then depth image quality can be improved, but this requires professional experience and increases operation complexity

Engineering Contradiction:
Improvedepth image qualityVSAvoidparameter setting complexity
Core Design Contradiction:
Manufacturing precisionVSDevice complexity

Solution Approach 1:

The system analyzes depth image statistics (depth distribution, confidence coefficient distribution) to determine scene type and automatically selects filtering parameters. This feedback mechanism continuously monitors image characteristics and adjusts parameters accordingly, eliminating the need for professional manual adjustment while maintaining optimal depth image quality

Inventive Principle:
Principle #23Feedback

Solution Approach 2:

The system pre-establishes multiple filtering parameter configurations corresponding to different scene types. Before actual filtering, the system determines the current scene type and selects the pre-configured parameters, avoiding the need for real-time manual adjustment and reducing operational complexity while ensuring high image quality

Inventive Principle:
Principle #10Preliminary action

3Ease of operation

If adaptive filtering is implemented by determining scene type automatically, then parameter setting becomes easier, but additional processing steps are required to determine scene type

Engineering Contradiction:
Improveparameter setting easeVSAvoidprocessing process complexity
Core Design Contradiction:
Ease of operationVSDevice complexity

Solution Approach 1:

The system combines scene type determination and filtering parameter selection into a single integrated process. The scene type determination uses simple statistical analysis of depth distribution and confidence coefficients, which are already computed during depth image acquisition. This merging eliminates the need for separate manual parameter setting steps while keeping the overall process simple and automated

Inventive Principle:
Principle #5Merging (Combining)

Data Source

PatentEP3764317B1Depth image processing method and apparatus, and electronic device
Publication Date: 2026.03.25 GUANGDONG OPPO MOBILE TELECOMMUNICATIONS CORP LTD
  • EP3764317B1 patent drawingFigure 1~2
  • EP3764317B1 patent drawingFigure 3~4
  • EP3764317B1 patent drawingFigure 5~7

AI summary

The present disclosure provides a depth image processing method and apparatus, and an electronic device. The method includes: acquiring a first image acquired by a depth sensor and a second image acquired by an image sensor; determining a scene type according to the first image and the second image; and performing a filtering process on the first image according to the scene type.