A multispectral image is one that captures image data from two or more ranges of frequencies along the spectrum, such as visible light and infrared energy. In multispectral images, the same spatial region is captured multiple times using different imaging modalities.

1. 10/23/2016 1
By
Alaa Mohammed Khattab

2. 
 Introduction.
 Electromagnetic Spectrum.
 Multi-Spectral Imaging.
 How multi-Spectral Imaging can be used.
 Multi-Spectral Cameras.
 Applications.
 Compression.
10/23/2016 2
Content

3. 
10/23/2016 3
Electromagnetic Spectrum

4. 
 Blue: 450-515..520 nm. Used for atmospheric and deep water
imaging.
 Green: 515..520-590..600 nm. Used for imaging of vegetation
and deep water structures.
 Red: 600..630-680..690 nm. Used for imaging of man-made
objects.
10/23/2016 4
Visible Light

5. 
 RGB which lie in the visible range can be combined
to produce a full color image for viewing on a
display.
 But, RGB are not sufficient to specify the appearance
of an object under varying illuminants and viewing
conditions.
10/23/2016 5
Visible Light

6. 
 Near infrared: 750-900 nm, is used primarily for
imaging of vegetation.
 Mid-infrared: 1550-1750 nm, is used for imaging
vegetation, soil moisture content, and some forest
fires.
 Far-infrared: 2080-2350 nm, is used for imaging soil,
moisture, geological features, silicates, clays, and
fires.
10/23/2016 6
Infrared

7. 
10/23/2016 7
Green vs. NIR
Green Reflectance NIR Reflectance

8. 
 A multispectral image is one that captures image
data from two or more ranges of frequencies along
the spectrum, such as visible light and infrared
energy.
 In multispectral images, the same spatial region is
captured multiple times using different imaging
modalities.
10/23/2016 8
Multi-Spectral Imaging

9. 
 Recording data from the visible spectrum showing
an object as it is seen by the eye.
 Take images in area of spectrum infrared region
(invisible).
10/23/2016 9
How can be used

10. 
Multi-Spectral Imaging
10/23/2016 10

11. 
 Multi-spectral systems make it possible to look
beyond those boundaries. Quest innovations is lead-
ing in multi-spectral camera technology.
10/23/2016 11
Multispectral Cameras

12. 
 Higher resolution pictures due to larger sensors
 Able to see in Wavelengths from 250 – 2200 nm
 Cameras ranging from 1 to 5 channels (a regular
camera is 1 channel)
10/23/2016 12
Multispectral Cameras

13. 
 More channels show several different information
layers
 Customized software to analyze every channel sepa-
rately and combined
 Compact and light systems (camera, information
storage, GPS/IMU integration)
10/23/2016 13
Multispectral Cameras

14. 
 Remote Sensing:
• The sensing platform is usually an aircraft or
satellite.
• The scene being imaged is usually the earth's
surface.
• The sensor and the target are so far apart, so each pixel
in the image can correspond to tens or even hundreds
of square meters on the ground.
10/23/2016 14
Applications

15. 
10/23/2016 15
Applications
Athens as seen by the SPOT 5
satellite in 2002
Spot-5 Satellite
French Spot

16. 
 American Thematic mapper sensors feature seven
bands of image data (three in visible wavelengths,
four in infrared)
10/23/2016 16
Applications

17. 
 VASARI SYSTEM:
• VASARI imaging system developed at the National
Gallery in London.
• we are investigating the use of seven filters spanning the
visible spectrum. These seven channels can be combined to
give accurate measurements.
10/23/2016 17
Applications

18. 
10/23/2016 18
Applications

19. 
 Interpretation of ancient papyri, such as those found
at Herculaneum, by imaging the fragments in the
infrared range (1000 nm).
 Used for document inspection and restoration as
hidden markings can be become visible in infrared.
10/23/2016 19
Applications

20. 
10/23/2016 20
Applications
Fingerprint Spoof Detection

21. 
10/23/2016 21
Applications
Extremely Dry Skin Multispectral
Fingerprint Spoof Detection

22. 
 Medical:
• Multispectral medical images can be formed from
different medical imaging modalities such as MRI, CT,
and X-ray. These multimodal images are useful for
identifying and diagnosing medical disorders.
10/23/2016 22
Applications

23. 
 For example ,The French SPOT and American
thematic mapper are older systems, use only a
handful of spectral bands.
 More modern systems, however, can incorporate
hundreds of spectral bands into a single image.
10/23/2016 23
Compression

24. 
 It is used to minimize transmission bandwidth from
the sensing platform to a ground station and to
archive the captured image.
 Steps:
• Transformation.
› Karhunen-Loeve (KL).
› Frequency transform.
• Encoding
10/23/2016 24
Compression

25. 
10/23/2016 25
Compression

26. 
 The optimal linear transformation for de-correlating
the data is the well-known Karhunen-Loeve (KL)
transform.
 Frequency transform such as the discrete cosine
(DCT) and wavelet transforms approach the optimal
KL transform point much more quickly than many
other transforms, so they are preferred in actual
compression systems.
10/23/2016 26
Transformation

27. 
 Multispectral data can be represented as a 2D image
with vector-valued pixels.
 Each pixel consists of one sample from each image
plane
 All spectral band:
• sampled at the same resolution.
• sampled over the same spatial extent.
• are perfectly registered, so each pixel component
corresponds to the same exact location in the scene.
10/23/2016 27
Transformation

28. 
 Transform-based multispectral coders is categorized
into three classes:
• Spectral-spatial transform.
• Spatial-spectral transform.
• Complex spatial-spectral transform.
10/23/2016 28
Transformation

29. 
 All images:
• Same spatial resolution.
• Registered.
10/23/2016 29
Spectral-spatial transform:

30. 
 Infrared band may have lower spatial resolution
than a visible band.
 All images:
• Not the same spatial resolution.
• Registered.
10/23/2016 30
Spatial-spectral transform:

31. 
 A spatial shift in one image plane relative to another.
 All images:
• Not the same spatial resolution.
• Not registered.
10/23/2016 31
Complex spatial-spectral transform:

32. 
 Multispectral image compression algorithms fall into
two general categories:
• Lossy.
• Lossless.
10/23/2016 32
Compression

33. 
Lossy algorithms typically obtain much
higher compression ratios but introduce
distortions in the decompressed image.
• Ex.:
› RGB color images.
› Remote sensed data.
10/23/2016 33
Lossy Compression

34. 
 The decoded image is identical to the original. This
gives perfect fidelity but limits the achievable
compression ratio.
• Ex.:
› Medical images.
› Photographic images.
10/23/2016 34
Lossless Compression

35. 
 Application using multispectral imaging for search:
• Fingerprint detection.
• Detection the originality of pages, paints,
documents.
• The discovery of forgery.
10/23/2016 35
Suggested Topics