Variable Kernel Morphological Operations for 3D Image Object Quality
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing image processing systems in video surveillance face issues with object separation and quality enhancement due to undesirable merging of objects and pixel artifacts, particularly when performing operations like dilate and erode on image data from 3D scenes.
Innovation Solution
The system identifies object pixels and additional pixels to associate with them, performing operations based on the depth and perspective of objects in the scene using a variable-sized kernel to improve object quality, employing a 'close' operation that adapts to the 3D perspective effects.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of manufacture
If fixed-kernel morphological operations (dilate and erode) are performed on image data, then object manipulation is achieved, but objects merge undesirably and pixel artifacts are created
Solution Approach 1:
The patent applies dynamics by transitioning from fixed-kernel to variable-kernel morphological operations. The kernel size and shape are dynamically adjusted based on the depth distance of objects from the camera, allowing the operation parameters to adapt to different spatial positions in the scene. This resolves the contradiction by maintaining operational simplicity while achieving precise object-quality preservation through depth-adaptive kernel selection.
Solution Approach 2:
The patent implements local quality by applying different morphological operation parameters to different regions of the image based on object depth. Objects at varying distances from the camera receive customized kernel treatments, with nearer objects using smaller kernels and farther objects using larger kernels. This localized adaptation prevents unwanted merging and artifacts while maintaining overall system simplicity.
2Productivity
If conventional morphological operations are used without depth consideration, then processing is simple, but object separation quality deteriorates
Solution Approach 1:
The patent applies parameter changes by modifying the morphological operation kernel parameters based on object depth distance. The system calculates the depth of each object from the camera and adjusts the kernel size and shape accordingly. This parameter adaptation maintains processing efficiency while significantly improving object separation accuracy, as the kernel parameters are optimized for each object's spatial position without requiring complex reprocessing.
3Manufacturing precision
If depth-based variable kernel operations are performed, then object quality improves and merging is prevented, but system complexity increases
Solution Approach 1:
The patent applies preliminary action by performing depth estimation and kernel parameter selection before executing the morphological operations. The system pre-calculates the appropriate kernel parameters based on object depth distance, allowing the actual morphological operation to proceed efficiently with predetermined parameters. This preliminary preparation reduces the complexity burden during the main processing stage while maintaining high object quality results.
Data Source
AI summary
Systems, methods, and software for operating an image processing system are provided herein. In a first example, a method of operating an image processing system is provided. The method includes identifying object pixels associated with an object of interest in a scene, identifying additional pixels to associate with the object of interest, and performing an operation based on a depth of the object in the scene on target pixels comprised of the object pixels and the additional pixels to change a quality of the object of interest.