Capturing information with imaging spectroscopy

Imaging is visualisation of an object. At specific spectral band it helps to visualise specific aspects of an object which are otherwise invisible to human eye. Spectroscopy captures information about specific aspects of an object without visualisation. Imaging spectroscopy is a merger of the potentials of both imaging and spectroscopy.

This blog is part of a series  and touches upon some background knowledge before exploring specific advanced sensing technologies, such as hyperspectral imaging.


The human eye is sensitive to red, green and blue wavelengths. The part of the electromagnetic spectrum that is visible to the human eye is for obvious reasons called ‘the visible spectrum’. The technological parallel to the human eye is a digital camera whose sensitivity is designed to peak in green wavelengths just like the human eye.

Red, green and blue image data are combined to form a digital colour image
(Source: Chem. Soc. Rev., 2014,43, 8200:

Visualising aspects otherwise invisible

Imaging at specific spectral band helps to visualise specific aspects of an object which are otherwise invisible to the human eye. Humans are practically blind outside the visible spectrum. However, some animals can visualise objects in wavelengths outside the visible spectrum. Examples are snakes, seeing heat signatures in the infrared (IR) spectrum, or bees, spotting nectar in ultraviolet (UV) spectrum. The parallel in the IR spectrum are the well-known thermographic cameras, which have a range of applications from low cost movement detectors to cooled thermal cameras that are used, for example, in testing the quality of automobile tires. As for UV, the examples range from low cost machine vision UV cameras used for inspection to more expensive scientific grade cameras.

Imaging is also not specific to a certain wavelength range. Some species such as, owls and sharks, do not detect specific wavelengths, meaning they are colour blind. Instead they have evolved to visualise intensity variations much like a grayscale camera.

A grayscale image data is collected from a single channel
(Source: Chem. Soc. Rev., 2014,43, 8200 :

Imaging systems can exploit different phenomena. Not all imaging systems are optical in nature. The optical region is UV – IR in the electromagnetic spectrum. And not all imaging systems are based on electromagnetic phenomena. Examples: positron emission tomography (PET) uses gamma rays, magnetic resonance imaging (MRI) uses magnetic fields and radio waves, ultrasound uses sound waves, certain microscopy techniques  are based on fluorescence detection, polarisation microscopy based on polarised light, and so on.


Spectroscopy is capturing information about an object without visualisation. Spectroscopy is very different from imaging in that it captures the information from a phenomenon of electromagnetic waves interacting with a compound at molecular and atomic levels without actual visualisation.

Titbit: the spectra that are absorbed or emitted by a compound are dispersed into its constituting wavelengths by a spectroscope. A spectroscope thus identifies the presence of a wavelength. The simplest example of a spectroscope is a prism.

Spectroscopy can identify pure compounds. The interaction of electromagnetic wavelengths with matter generates a response that is the same for identical molecules or atoms. This response is referred to as spectral signature of the (chemical) species. Any variation in the molecular or atomic composition of a material will generate a response with the corresponding spectral signatures. Thus, using spectral signatures compounds can be uniquely identified, making spectroscopy a useful tool in qualitative and quantitative measurements of compounds.

Spectroscopy at specific spectral band provides information about specific aspects of an object. In spectroscopy, similar to imaging, the unique identifiers of a certain feature could be present in a certain part of the electromagnetic spectrum and absent in others. For example, infrared spectroscopy is a rather suitable technology for agricultural application such as crop disease, crop species identification, due to infrared being able to detect vibrations of the bonds in organic compounds.

Typical application areas

Typical application areas for spectroscopy are agriculture, food, bio-actives, pharmaceuticals, petrochemicals, textiles, cosmetics, medical applications, and chemicals such as polymers.

To recycle plastics HSI in NIR (1 000 - 1 700 nm) and SWIR (1 000 - 2 500 nm) wavelengths is useful. To sort black plastics MWIR (3 000 - 5 000 nm) wavelength can be utilised.

Imaging spectroscopy

Imaging spectroscopy is essentially a merger of the potentials of both imaging and spectroscopy, thereby having the ability to visualise materials and determine their characteristics, i.e. provide the spatial as well as spectral information.

Imaging, Spectroscopy and imaging spectroscopy
(Source: Mehta N, Shaik S, Devireddy R, Gartia M. Single-Cell Analysis Using Hyperspectral Imaging Modalities. ASME. J Biomech Eng. 2018;140(2):020802-020802-16. doi:10.1115/1.4038638)

Imaging spectroscopy provides spectral signature at every pixel. Simply put, the object of interest is imaged at spectral bands of interest, implying that every pixel of the image contains a spectral signature.

Unlike spectroscopy, this technique does not provide a mean spectral signature of the material, rather it provides the spectral signature at every pixel area. Unlike RGB images, the individual images at each spectral band are not automatically merged to form a single colourful image. Therefore, the images are separately available as respective monochrome images.

Imaging spectroscopy generates more information at larger acquisition time than a RGB image. Unlike an RGB image, information is not lost to interpolation as the separate images are not automatically merged to a single colour image.

Four generations

Generally, imaging spectroscopy can be classified into four generations: multispectral, hyperspectral, ultraspectral and fullspectral imaging, which we compare in the table below. Multispectral imaging is performed with fewer number of spectral bands, typically up to 10 bands. Hyperspectral imaging is of finer resolution where imaging is done at hundreds of spectral bands and these spectral bands are quite narrow, i.e., below 10 nm of full width at half maximum.

From RGB to multispectral to hyperspectral, there is clearly an increase in the number of spectral bands and subsequent channels required to image a material. The image can be viewed as 3D data with spatial and spectral axes, This data cube increases in size as the number of spectral bands and spatial extent increase. For a given hardware setup, an increase of the data cube results in a longer image acquisition time.


The AVIRIS image cube shows 224 spectral channels stacked on each other, with the uppermost slice imaged in visible spectrum
(Source: NASA/JPL-Caltech)

Classification of imaging spectroscopy 

Application areas 

Typical application areas for imaging spectroscopy are: 

Multispectral: food/agriculture, pharmaceuticals, geology/atmosphere, … 

Hyperspectral: material characterisation, recycling, biometrics, remote sensing, chemometrics, environmental monitoring, forensics and counterfeits, …  

In the next article we will delve further into the developments in hyperspectral imaging and its applications. 

Want to know more about the use of sensors in food industry? On 22 November we will be organising an event on advanced sensors within the project SensInFood. You can find further information on this event here. This event is in cooperation with Flanders' Food and with the support of VLAIO.