Joint Image Compression Encryption for IoT
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing image compression techniques are inefficient for IoT applications, as they prioritize human-perceivable visual quality over machine-to-machine communication and resource-constrained environments, leading to high communication bandwidth and energy consumption.
Innovation Solution
A machine learning-guided joint compression and encryption method that utilizes a color quantization dictionary and common pattern dictionary to segment images into triangles, allowing for high compression ratios while maintaining accuracy for coarse-grained vision tasks, and optionally encrypts images using a secret dictionary.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional image compression techniques are used, then visual quality for human perception is maintained, but communication bandwidth and energy consumption increase significantly
Solution Approach 1:
The patent segments images into triangular meshes and compresses them by encoding vertex positions and color information separately. This segmentation allows the system to prioritize encoding only the essential geometric and color data needed for machine vision tasks rather than preserving all visual details, thereby reducing energy consumption while maintaining sufficient quality for automated analysis.
Solution Approach 2:
The patent applies different compression strategies to different regions of the image based on their importance for machine vision tasks. Critical regions containing features relevant to automated analysis are preserved with higher fidelity, while less important regions are compressed more aggressively, optimizing the balance between quality and energy efficiency for specific application needs.
2Measurement precision
If traditional image compression techniques are used, then visual quality for human perception is maintained, but communication bandwidth increases
Solution Approach 1:
By segmenting images into triangular meshes and encoding only essential vertex positions and color information, the patent dramatically reduces the data volume transmitted over communication channels. This segmentation approach eliminates redundant pixel information while preserving the geometric and visual features necessary for machine vision applications, thereby reducing bandwidth requirements.
Solution Approach 2:
The patent extracts and transmits only the essential features needed for machine vision tasks—specifically vertex positions, normal vectors, and color information of triangular meshes—rather than transmitting complete pixel data. This extraction of critical information significantly reduces communication bandwidth while maintaining sufficient quality for automated analysis.
3Productivity
If compression ratio is increased for IoT applications, then energy and transmission efficiency improve, but accuracy for vision tasks may deteriorate
Solution Approach 1:
The patent applies differential compression quality to different image regions based on their importance for machine vision tasks. Regions containing critical features for automated analysis are preserved with higher accuracy, while less important regions undergo more aggressive compression, thereby maintaining transmission efficiency while preserving necessary accuracy for vision tasks.
Solution Approach 2:
The patent incorporates feedback mechanisms where the compression process is guided by the specific requirements of machine vision tasks. The system adjusts compression parameters based on the detected importance of different image regions for the intended application, ensuring that accuracy is maintained where needed while maximizing compression where possible.
4Reliability
If separate compression and encryption processes are used, then security is ensured, but overhead and operational costs increase
Solution Approach 1:
The patent merges compression and encryption into a unified process by performing encryption on the compressed triangular mesh data rather than on the original image. This combination eliminates the need for separate compression and encryption operations, reducing computational overhead and operational costs while maintaining both compression efficiency and security.
Solution Approach 2:
The patent performs compression before encryption, preparing the data in advance by reducing its size and structure. This preliminary compression action reduces the amount of data that needs to be encrypted and processed, thereby reducing the overhead and computational requirements of the subsequent encryption operation while maintaining security.
Data Source
AI summary
A joint compression and encryption system is configured to retrieve, from a local memory or image capture device, an image file including a first plurality of segments. The system is further configured to identify, for one or more segment of the first plurality of segments, a matching segment in a local segmentation repository. The system is further configured to compress remaining segments of the first plurality of segments for which no matching segment was identified into a compressed remaining segment set. The system is further configured to transmit, via an unsecure communication channel and to a second computing entity, the identifications of the matching segments, and the compressed remaining segment set. The system is further configured to, prior to retrieving the image file, perform a handshake or calibration process. The system is further configured to encrypt the matching segments using a key into an encrypted segment set and transmit the key to the second computing entity. In embodiments, the second computing entity, using the key, decrypts the encrypted segment set.