LightForm_Logo
Copyright © LightForm Inc, 2017
LightForm Inc: Pioneering Analytical Hyperspectral Microscopy Since 1996
The Pain: Linear algorithms are based on assumptions that often fail at the micro level:  Discrete and non-interacting objects: If spectral objects in the FOV are discrete and non-interacting, then spectral mixing is simply the sum of all spectra present within the spatial resolution of the instrument. (Keshava and Mustard*).  This is almost never the case in natural samples at a micro level can can result in inaccurate spectral segmentation. Active illumination: The illuminating light source, often a laser, actively induces change in the spectral characteristics of objects in the FOV.  Linear algorithms demands passive illumination of the FOV Instrumental factors: linear mixing depends on the spatial resolution of the instrument system (microscope, telescope, slit width, pixel size...).  “Zooming in” should reveal the pure spectrum of each object.  This is either impossible or unlikely with natural samples at a micro level,  LightForm rejects the use of linear un-mixing algorithms:  In most Hyperspectral characterizations of natural or biological material, objects are rarely truly discrete and non-interacting.  Conditions that present light scatter, photo-induced reactions, changes in ion/pH, conformation.

Hyperspectral Linear Algorithms Algorithms Fail

Linear Algorithms Often Fails at the Microscopic Level
* For literature references go here
Literature Refs Literature Refs
Hyperspectral un-mixing algorithms:  Originated in the remote Earth Sensing community. Due to the success and future promise of this technology a considerable amount of effort has been devoted to developing software to process and generate Hyperspectral Images.  Commercial software development: In order to economically serve this community commercial programs are available typically for spectral analysis and linear spectral un-mixing. Justification: Most fields of view in remote sensing are non-interacting, passively reflect light, and are spatially separated.  (For example the spectrum of a brick may mix with that of mortar from a distance but are otherwise discrete) The microscopy dilemma: None of these benign conditions are likely to exist in biological materials or indeed most samples  at a microscopic level. Consequently, linear algorithms often fail to produce accurate results.
LightForm_Logo
Copyright © LightForm Inc, 2017
LightForm Inc: Pioneering Analytical Hyperspectral Microscopy Since 1996
The Pain: Linear algorithms are based on assumptions that often fail at the micro level:  Discrete and non-interacting objects: If spectral objects in the FOV are discrete and non-interacting, then spectral mixing is simply the sum of all spectra present within the spatial resolution of the instrument. (Keshava and Mustard*).  This is almost never the case in natural samples at a micro level can can result in inaccurate spectral segmentation. Active illumination: The illuminating light source, often a laser, actively induces change in the spectral characteristics of objects in the FOV.  Linear algorithms demands passive illumination of the FOV Instrumental factors: linear mixing depends on the spatial resolution of the instrument system (microscope, telescope, slit width, pixel size...).  “Zooming in” should reveal the pure spectrum of each object.  This is either impossible or unlikely with natural samples at a micro level,  LightForm rejects the use of linear un-mixing algorithms:  In most Hyperspectral characterizations of natural or biological material, objects are rarely truly discrete and non-interacting.  Conditions that present light scatter, photo-induced reactions, changes in ion/pH, conformation.

Hyperspectral “Un-Mixing” Software:

There Can Be Pain!

Linear Spectral Un-Mixing Often Fails at the Microscopic Level
* For literature references go here
Literature Refs Literature Refs
Hyperspectral un-mixing algorithms:  Originated in the remote Earth Sensing community. Due to the success and future promise of this technology a considerable amount of effort has been devoted to developing software to process and generate Hyperspectral Images.  Commercial software development: In order to economically serve this community commercial programs are available typically for spectral analysis and linear spectral un- mixing. Justification: Most fields of view in remote sensing are non- interacting, passively reflect light, and are spatially separated.  (For example the spectrum of a brick may mix with that of mortar from a distance but are otherwise discrete) The microscopy dilemma: None of these benign conditions are likely to exist in biological materials or indeed most samples  at a microscopic level. Consequently, linear algorithms often fail to produce accurate results.