The analysis of image complexity has been around for over a decade, and numerous examples are available in the present blog. An image is an N x M matrix of pixels and is easy to analyze with, for example, OntoNet, the QCM engine from Ontonix. The first step is to transform an image to greyscale and then to determine the intensity of each pixel on scale ranging, for example, from 0 to 255.
There is a problem, however. If the image is rotated, even slightly, the result changes. Imagine, for example, that we are dealing with MRI images and suppose that two different images, taken at different times, are available, such as the ones below.
If the goal is to identify small local changes, or even large overall differences, this will surely pose a problem. While image complexity is invariant to translations, this is certainly not the case for rotations.
To obviate the problem, a new technique has been developed, which is based on specific mathematical transformations of the original image. The computational requirements are evidently higher than in the traditional approach. The new approach works for all sorts of images, including satellite imagery, infrared, etc.
The new Image Complexity Analysis is available only as a service.
Artificial Intuition 
0 comments on “Ontonix Develops Rotation-Invariant Image Complexity Analysis”