Depth-Aware Camera Image Stabilization Through Posture Reprojection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Traditional anti-jitter solutions for image stabilization assume all objects are on a unit plane, leading to unstable effects, and existing methods fail to effectively stabilize images due to external shaking during shooting.
Innovation Solution
An image processing method that involves acquiring current posture information, converting it into target posture information through low-pass filtering, and using depth information fusion to reproject images, incorporating gyroscope data and multiple depth sensors for accurate stabilization.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Device complexity
If traditional simplified processing is used to stabilize shooting images, then the processing complexity is reduced, but the anti-jitter effect becomes unstable
Solution Approach 1:
The patent changes the parameter assumption from simplified unit plane to actual depth information, transforming the processing model to achieve both stability and accuracy in anti-jitter effects
Solution Approach 2:
The patent replaces simplified geometric processing with physics-based depth sensing and coordinate transformation algorithms, substituting mechanical assumptions with sensor-driven data processing
2Device complexity
If all objects are assumed to be on a unit plane, then the processing model is simplified, but the depth accuracy of pixel points is reduced
Solution Approach 1:
The patent transitions from 2D unit plane assumption to 3D spatial coordinate system by incorporating depth dimension, enabling accurate depth measurement and three-dimensional coordinate transformation for each pixel point
Solution Approach 2:
The patent introduces depth sensors as intermediary devices to obtain actual depth information, serving as a bridge between the image plane and three-dimensional space
3Measurement precision
If multiple depth information sources are fused, then the depth accuracy is improved, but the processing complexity increases
Solution Approach 1:
The patent merges multiple depth information sources (depth sensor, dual-camera parallax, phase focusing) into a unified depth map through fusion processing, combining advantages of different methods to achieve superior depth accuracy
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 method achieves stable image stabilization by predicting and correcting camera jitter, enhancing anti-jitter performance and improving image quality by accurately determining pixel depth and posture adjustments.
Implementation Method 1
acquiring the current angular velocity information through a gyroscope
Implementation Method 2
information in the direction in which the jitter needs to be eliminated in the current posture information is subjected to the low-pass filtering process to remove a high-frequency component in the information
Implementation Method 3
the initial depth information is depth information of each pixel point in the image to be processed under the current posture acquired by a depth camera
Implementation Method 4
the third depth information is depth information of each pixel point in the image to be processed under the current posture acquired by a dual-camera parallax ranging method
Implementation Method 5
the fourth depth information is depth information of each pixel point in the image to be processed under the current posture acquired by a phase focusing method
Data Source
Figure 1
Figure 2
Figure 3~4
AI summary
Disclosed is an image processing method, comprising: converting current posture information of a camera into target posture information, the current posture information comprising current angular velocity information; acquiring first depth information of each pixel point in an image to be processed in a current posture; determining second depth information of each pixel point in a target posture according to the first depth information; and obtaining a target image according to the current posture information, the first depth information, the target posture information, the second depth information, and first internal reference information of the camera.