Adaptive Image Resolution Reduction for Surveillance Object Detection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing video surveillance systems face challenges in processing video sequences in real time due to high processor load, particularly in object segmentation and tracking, which is exacerbated by uniform resolution reduction across images, leading to poor detection of remote objects.
Innovation Solution
The method adjusts resolution reduction based on object distance in three-dimensional scenes, ensuring that regions with greater object distance maintain higher resolution, while regions closer to the camera undergo more significant reduction, using techniques like morphing and warping to maintain a similar number of image points for objects of equal size, regardless of their distance.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If uniform resolution reduction is applied across the entire image, then the processor load is reduced and productivity is improved, but the detection precision of remote objects deteriorates
Solution Approach 1:
The patent applies different resolution reduction factors to different regions of the image based on object distance. Remote regions undergo less resolution reduction than near regions, maintaining detection precision for distant objects while still achieving overall processor load reduction through selective downsampling of closer regions.
2Device complexity
If resolution is reduced equally in horizontal and vertical directions, then the device complexity is minimized and ease of manufacture is improved, but the detection precision of objects at different distances deteriorates
Solution Approach 1:
The patent implements dynamic resolution reduction where the reduction factor varies across different spatial regions and object distances. This dynamic approach allows the system to adapt the resolution reduction to the specific scene content and object positions, maintaining detection precision while managing device complexity through algorithmic flexibility rather than hardware complexity.
3Productivity
If uniform downsampling is applied to the entire video sequence, then the productivity is improved by reducing computing effort, but the reliability of object detection for remote objects deteriorates
Solution Approach 1:
The patent applies local quality by using different downsampling factors for different spatial regions corresponding to different object distances. Regions with remote objects use smaller downsampling factors to maintain detection reliability, while regions with near objects use larger downsampling factors to reduce computing effort, achieving an optimized balance between productivity and reliability.
Data Source
AI summary
In a method, a device, and a computer program for reducing the resolution of an input image, wherein input image shows a three-dimensional scene that was recorded by a surveillance camera, the distance between surveillance camera and objects in the three-dimensional scene is referred to as the object distance, and the resolution reduction is adjusted with consideration for the object distance in the three-dimensional scene and/or the perspective distortion of the three-dimensional scene.