| Author | Message | ||
     Bruce H. Campbell (Campbell) |
With the increase in the use of imaging, does anyone know if chemometrics is used to extract information from the image? If chemometrics is used, what is its form? Similar to the ones normally used for chemometric near IR? | ||
| Tony Davies (Davies) |
The answer is "Yes". Paul Geladi has written a book (available from NIR Publications!) Multivariate Image Analysis. It is considering optical images but much of the chemometrics would be directly transferable to NIR images. Lots of PCA! | ||
| Bruce H. Campbell (Campbell) |
I was wondering if you, or someone else, could explain a little more on how the PCA works. Is PCA applied on each picture element? Is the result a quantitative prediction or is it qualitative? | ||
| paul geladi (Geladi) |
We have to make a dstinction. For univariate (e.g. 512x512) images, PCA can be used for compression by representing smaller blocks by a few components. For multivariate images (e.g. 512x512 x 10 wavelengths), latent variable (score) images can be made by calculating loadings for the 10 wavelengths. Score images remove noise and concentrate information in the first scores. They become even more interesting by making score plots and doing segmentation in score space. |