Image multiset analysis

Two examples will illustrate the advantages of multiset analysis in terms of the relevance of spectral consistency (and, therefore, definition of compounds) and on the advantages of using hyperspectral imaging toward classical spectroscopy in the modeling of complex processes (transformations). 

In the multiset analysis, spectral consistency and, therefore, compound identity are ensured because of the architecture of the multiset (with a single matrix), and distribution maps are better defined. The deficient quality of the results obtained by individual single image analysis in this example is due to two different factors (a) not all sample constituents are sufficiently represented in all emulsion layers and (b) the quality of the Raman signal (signal-to-noise ratio) decreases as depth increases. These problems are solved with multiset analysis because of the complementary composition information from the different images and the better signal-to-noise ratio of top emulsion layers that influence the quality of the overall multiset structure results. 

---

Figure 2.13 Resolution analysis of a multilayer emulsion image, (a) Distribution maps of emulsion layers and resolved pure spectra from top and bottom layers obtained by individual layered resolution and (b) pure spectra and distribution maps coming from image multiset analysis.

Figure 2.18 Graphical scheme of the combination of image multiset analysis and superres-olution postprocessing. (Reprinted from Ref. [20].)...

Absolute quantitative information is a concept linked to multiset image analysis and requires, as calibration methods do, the presence of a set of standard images with known composition and a set of unknown images, the concentration ofwhich wants to be predicted. 

The configuration of the multiset used for the analysis may also affect the quality of the final results. Two aspects should be taken into account the kind of information introduced in the multiset, that is, the composition of the calibration images used and the architecture of the multisets used for the calibration/validation steps. 

As a general conclusion, the strategy combining multiset image analysis and superresolution postprocessing simultaneously saves computation time and, above all, provides unmixed, spatially detailed, constituent-specific interpretable information on the sample analyzed. 

---