Multivariate Image Analysis

Multivariate Image Analysis (MIA), including Multivariate Image Regression (MIR) and Multivariate Texture Analysis (Angle Measure Technique, AMT) are ACABS new priority R&D application areas (2008).

MIA and MIR deal with multivariate and hyperspectral images (typically in visual light and NIR wavelength ranges) in which each pixel is a full spectrum. Chemometric projection (PCA) and regression approaches (PLS) allows to discriminate or classify pixels according to embedded features in depicted image objects and predicting quantitative or qualitative functional properties. While MIA is still a relatively new approach (one of the first MIA articles was published in 1989), it has recently seen rapidly growing application in the wide range of scientific, technological and industrial areas in chemical imaging, process monitoring, remote sensing, geological surveys, food industry a.o.

Multivariate texture analysis works on one-channel (grey-level) imagery extracting complexity and texture related information, typically using it further as input to prediction (PLS). ACABS has developed a new textural feature-extracting approach, based on the Angle Measure Technique (AMT), which is used extensively for powder characterization, analysis of particle size distribution, classification of blood cells and many others applications.