Thermal Infrared Image Compression for Stable Machine Vision Detection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing object detection technologies face challenges in efficiently utilizing thermal infrared images for machine vision due to inefficient encoding methods and the burden of high-quality images, especially in low-light environments, which can lead to unstable transmission and prolonged processing times.
Innovation Solution
A method and apparatus for compressing thermal infrared images by generating foreground and background images, adjusting bounding box sizes, and encoding these images with different quantization parameters to optimize compression and decoding for efficient object detection.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If high-quality images are used for object detection, then detection performance is improved, but transmission stability and processing efficiency deteriorate
Solution Approach 1:
The image is divided into multiple tiles or blocks, and each tile is processed and transmitted separately. This segmentation allows for more efficient compression and transmission of large images while maintaining detection accuracy in critical regions.
Solution Approach 2:
Different quantization parameters are applied to different regions of the image based on their importance. Regions containing objects of interest use lower quantization parameters (higher quality), while background regions use higher quantization parameters (lower quality), optimizing both detection performance and transmission efficiency.
2Measurement precision
If high-quality images are used for object detection, then detection performance is improved, but processing time increases
Solution Approach 1:
The image processing is divided into multiple parallel operations on different tiles, allowing for concurrent processing that reduces overall processing time while maintaining quality where needed.
Solution Approach 2:
By applying different quality levels to different regions, the system reduces the total amount of data that needs to be processed at high quality, thereby reducing processing time while preserving detection accuracy in critical areas.
3Ease of operation
If existing encoding methods are used for thermal infrared images, then transmission is simplified, but object detection accuracy deteriorates
Solution Approach 1:
Different quantization parameters are applied to different regions of the thermal infrared image based on object presence and importance. This region-dependent encoding maintains detection accuracy in object regions while simplifying encoding in background regions.
Solution Approach 2:
The quantization parameters are dynamically adjusted based on the detected objects and their locations in the image, allowing the encoding process to adapt to the specific content of each image rather than using a fixed encoding scheme.
Data Source
AI summary
Disclosed herein is a method for compressing an image for machine vision, the method including detecting objects in an input image using an object detection network, generating a foreground image including bounding boxes corresponding to the objects and a background image, which is an image acquired by excluding the bounding boxes from the input image, encoding the foreground image and the background image, and decoding the encoded foreground image and the encoded background image.