Optical Flow SGM Processing with Denoising and Edge Detection

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Existing methods for determining optical flow, such as conventional Semi-Global Matching (SGM), are inefficient and lack accuracy in processing image pairs, particularly in applications requiring precise motion estimation and object detection.

Innovation Solution

A modified Semi-Global Matching (SGM) method is implemented using optical flow hardware that includes denoising, edge detection, object detection, and decision logic processes to generate accurate disparity maps, enhancing the precision of optical flow calculations.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If conventional Semi-Global Matching (SGM) is used to determine optical flow, then the processing speed is maintained at acceptable levels, but the accuracy and precision of motion estimation deteriorate

Engineering Contradiction:

Improveoptical flow accuracyVSAvoidprocessing efficiency

Core Design Contradiction:

Measurement precisionVSProductivity

Solution Approach 1:

The patent divides the image processing into distinct stages: denoising, edge detection, object detection, and decision logic. Each stage processes specific aspects of the image data independently, allowing conventional SGM to operate on pre-processed, enhanced data that improves accuracy without requiring complete redesign of the matching algorithm itself.

Inventive Principle:

Principle #1Segmentation

Solution Approach 2:

The patent applies preliminary processing steps (denoising, edge detection, object detection) before the actual optical flow calculation. These preliminary actions prepare the image data by removing noise, identifying important features, and segmenting objects, which then enables more accurate matching results from the conventional SGM algorithm without sacrificing processing speed.

Inventive Principle:

Principle #10Preliminary action

2Reliability

If conventional SGM is used for optical flow calculation, then computational resources are used efficiently, but measurement precision and reliability deteriorate

Engineering Contradiction:

Improvemotion estimation reliabilityVSAvoidcomputational resource efficiency

Core Design Contradiction:

ReliabilityVSUse of energy by moving object

Solution Approach 1:

The patent applies different processing qualities to different regions of the image. Edge detection and object detection are performed selectively based on image content, with more computational resources allocated to regions containing important features or objects, while simpler processing is applied to homogeneous regions. This local quality approach improves reliability without uniformly increasing computational cost across the entire image.

Inventive Principle:

Principle #3Local quality

Solution Approach 2:

The patent introduces intermediary processing layers (denoising module, edge detection module, object detection module) that act as mediators between the input image and the optical flow calculation. These intermediaries preprocess the data to improve quality and reliability, filtering out noise and highlighting important features before the final matching operation, thereby improving reliability with acceptable computational overhead.

Inventive Principle:

Principle #24Intermediary (Mediator)

Data Source

PatentUS20250315963A1Computing optical flow using semi-global matching

Publication Date: 2025.10.09 NVIDIA CORP

AI summary

Apparatuses, systems, and techniques to determine optical flow. In at least one embodiment, a set of disparity values is used to determine optical flow between input and reference images. For each of a plurality of image regions of the input image, the set of disparity values may include disparity values for a plurality of directions intersecting the image region.