Research
We focus on comprehensive exploration of our technology. Its efficiency has been proved in term of various performance indicators. And new features are discovered.
A special class of atomic functions is a base of the developed methods. They have useful constructive properties, and it is their applying that ensures a combination of data encryption and compression features with machine learning oriented image representation. Also, compactness of the applied functions ensures low resource intensive processing. In order to provide fundamental base and scientific validation, we investigate efficiency of the created algorithms in terms of image analysis, processing, encryption, complexity analysis and with respect to various performance indicators. There are two principal results of our research: atomic functions are perfect for image processing and
it is just the beginning!
Encryption
Base
constructive properties of the applied atomic functions
Properties
-
it's a built-in feature
-
correct reconstruction requires a correct key
-
there are more than 1E+176 encryption keys
Compression
Base
approximative properties of the applied atomic functions
Features
-
lossy & lossless modes
-
classic, block-splitting & chroma subsampling modes
-
compression variation mechanism
-
quality control (MAD, RMSE & PSNR metrics)
Low Complexity
Base
compactness of the applied atomic functions
Features
-
T(n) = O(n), i.e. linear time complexity (n - size of the data processed)
-
S(n) = O(n), i.e. linear spatial complexity (classic & chroma-subsampling modes)
-
S(n) = O(1), i.e. constant spatial complexity (block-splitting mode)
ML (in progress)
Base
-
approximative properties & compactness of the applied atomic functions
-
rich image view representation
-
AI/ML-oriented data format
Goals
-
low resource intensive computer vision methods development
-
outperform the existing algorithms
