Dynamic Object Template Update for Video Segmentation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional video segmentation methods face challenges in accurately segmenting foreground objects due to the need for a pre-determined static background image, which requires no moving objects and can be cumbersome to regenerate if the camera position changes, and are prone to noise and undesired holes in depth-based methods.
Innovation Solution
A video processing system that dynamically updates object templates using a mean-shift technique to detect and segment objects in real-time, selecting the best template based on color values and position, and selectively using depth values from a depth sensor for reliable regions.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of manufacture
If a pre-determined static background image is used for foreground segmentation, then the segmentation process is simple, but the system becomes cumbersome to regenerate when camera position changes and cannot handle moving objects in the background
Solution Approach 1:
The patent transitions from a static background image to a dynamic background model that is continuously updated using the mean-shift technique. The system dynamically adjusts the background model based on current image data, allowing it to adapt to camera position changes and moving objects without requiring manual regeneration of the background image.
Solution Approach 2:
The system implements feedback by continuously comparing the current image with the updated background model and using the mean-shift technique to refine the background representation. This feedback loop allows the background model to learn from and adapt to changing conditions in real-time.
2Adaptability or versatility
If depth information from a depth sensor is used for foreground segmentation, then the system can handle moving objects, but the boundaries become noisy and contain undesired holes due to severe noise in depth sensors
Solution Approach 1:
The patent changes the approach from relying on depth values to using color information and the mean-shift technique. By parameterizing the background model in terms of color distribution and using iterative mean-shift optimization, the system achieves smooth boundaries without the noise problems associated with depth sensor data.
Solution Approach 2:
The system replaces the mechanical/physical depth sensing mechanism with a computational approach using color image processing and the mean-shift algorithm. This substitution eliminates the noise issues inherent in depth sensors while maintaining the ability to segment moving objects.
3Device complexity
If a completely static background image is generated, then the segmentation process is straightforward, but the system cannot accommodate objects that move in the background scene
Solution Approach 1:
The system transforms the background model from a completely static image to a dynamic representation that can adapt to moving objects. The mean-shift technique enables the background model to dynamically adjust and accommodate objects that move through the scene, improving segmentation accuracy without significantly increasing complexity.
Data Source
AI summary
Various aspects of a video-processing system and method for object detection in a sequence of image frames are disclosed herein. The system includes an image-processing device configured to receive a first object template for an object in a first image frame that includes one or more objects. A plurality of object candidates that corresponds to the object for a second image frame are determined by use of the shape of the received first object template. One of the determined plurality of object candidates is selected as a second object template, based on one or more parameters. The received first object template is updated to the selected second object template to enable segmentation of the object in the second image frame and/or subsequent image frames.