Monoscopic Depth Determination via Motion Parallax and ML
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Determining depth values for monoscopic images captured with a moving camera is challenging due to the difficulty in distinguishing between camera motion and object movement, making it hard to apply triangulation methods effectively.
Innovation Solution
A machine learning model is configured to determine depth values by combining information from a target image and a reference image, using an object mask to separate static and moving features, and generating a dynamic depth image that represents depth values for both static and moving features based on motion parallax and camera pose.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If triangulation is applied to monoscopic images captured with a moving camera, then depth values can be determined, but it becomes difficult to distinguish between camera motion and object movement
Solution Approach 1:
The patent segments the image into static and moving regions by comparing the target image with a reference image. By identifying regions that remain unchanged (static) versus regions that change (moving), the system can apply different depth determination strategies. Static regions use traditional triangulation while moving regions are handled separately, resolving the ambiguity of camera motion versus object movement.
Solution Approach 2:
The patent introduces an intermediary reference image that serves as a mediator between the target image and the depth determination process. This reference image captures the scene at a previous time point and allows the system to identify moving objects by comparing differences, thereby separating camera-induced motion from object-induced motion.
2Measurement precision
If stereoscopic cameras are used to capture images in parallel, then depth values for both moving and static features can be determined, but the device complexity increases
Solution Approach 1:
The patent creates a virtual stereoscopic effect by generating a synthetic reference image from the monoscopic target image and previously captured frames. Instead of requiring physical stereoscopic cameras, the system copies and processes temporal information to simulate the depth information that would be obtained from parallel camera captures, thereby maintaining depth determination capability while avoiding increased hardware complexity.
Solution Approach 2:
The patent replaces the mechanical stereoscopic camera system with a computational approach. Instead of using physical parallel cameras to capture depth information, the system uses image processing algorithms and temporal comparison to synthesize depth data, substituting mechanical complexity with computational processing.
Applied Scientific Principles
This section explains which scientific principles are used to turn an abstract innovation direction into a practical engineering solution.
Function Achieved in This Case
The solution enables accurate determination of depth values for both static and moving features in monoscopic images, improving the ability to generate depth maps and apply depth-based visual effects, even when stereoscopic images are unavailable.
Implementation Method 1
determining, by the processor and based on motion parallax between the reference image and the target image, a static depth image that represents depth values of the static features in the target image
Data Source
AI summary
A method includes obtaining a reference image and a target image each representing an environment containing moving features and static features. The method also includes determining an object mask configured to mask out the moving features and preserves the static features in the target image. The method additionally includes determining, based on motion parallax between the reference image and the target image, a static depth image representing depth values of the static features in the target image. The method further includes generating, by way of a machine learning model, a dynamic depth image representing depth values of both the static features and the moving features in the target image. The model is trained to generate the dynamic depth image by determining depth values of at least the moving features based on the target image, the object mask, and the static depth image.