Organizing a spectral image database by using SelfOrganizing Maps

1
Organizing a spectral image database by using
Self-Organizing Maps
Research Seminar 7.10.2005 Oili Kohonen
2
Motivation?

• Image retrieval from conventional databases
  since 1990's ... many efficient
  techniques have been developed
• However, efficient techniques for querying
  images from spectral image
  database does not exist.
• Due to the high amount of data in the case of
  spectral images, the efficient techniques
  will be needed.

3
Spectral imaging?

Metameric imaging cheap and practical way to
achieve a color
match. Spectral imaging needed
to achieve a color match for all
observers across
the changes in the illumination.
4
Principle of SOM

• The Self-Organizing Map (SOM) algorithm
• Is an unsupervised learning algorithm.
• Defines mapping from high-dimensional data into
  lower-dimensional data.
• SOM
• Consists of arranged units (or neurons), which
  are represented by weight vectors.
• Units are connected to each other by neighborhood
  relation.

5
Principle of SOM

• SOM Algorithm
• begin
• Initialize the SOM
• for i 1 number of epochs
• take input vector x randomly from the training
  data
• find the BMU for x
• update the weight vectors of the map
• decrease the learning rate neighborhood
  function
• end
• end

6
Principle of SOM finding the BMU
Euclidean distance is a typically used distance
measure.
7
Principle of SOM updating the weight vectors
Learning rate product of learning rate parameter
neighborhood function
8
Principle of SOM neighborhood function

• 
  Neighborhood function
• h(t) has to fullfill the following two
  requirements
• It has to be symmetric about the maximum
  point (BMU).
• It's amplitude has to decrease monotonically
  with an increasing distance from BMU.

9
Principle of SOM Lattice structure
Lattice structures hexagonal rectangular
10
Searching Technique Constructing histogram
database

• Train SOM
• Find BMU for each pixel in an image
• Generate BMU-histogram normalize it by the
  number of pixels in an image
• Repeat steps 2 3 for all images in a spectral
  image database
• Save histogram database with the information of
  SOM-map

11
Searching Technique making a search

• Choose an image and generate its histogram.
• Calculate the distances between the generated
  histogram and the existing histogram
  database.
• Order images by these distances.

The results of the search are shown to user as
RGB-images
12
Searching techniques
One-dimensional SOM
13
Searching techniques
Two-dimensional histogram-trained SOM
14
Distance Calculations
15
Experiments
(Unweighted images)
(Unweighted and weighted images)
16
The Used Database
17
Training of the SOMs

• 10 000 spectra were selected randomly from each
  image.
• 2 000 000 4 000 000 epochs in ordering fine
  tuning phases, respectively.
• Unit sizes 50 chosen empirically
  
  49 to have comparable results with 1D-SOM
  1414 map in the case of
  histogram-trained
  SOM

18
Results 1d-SOM, Unweighted images
Multiplied data
Pure data
The distance measure Euclidean distance
19
Results 1D, Unweighted images

• Energy
• K-L
• Peak
• DPD
• JD

20
Results 1D, Weighted images

• Energy
• K-L
• Peak
• DPD
• JD

21
Conclusions I

• The structure of the database is different for
  weighted and unweighted images.
• The best results were got by using euclidean
  distance and Jeffrey divergence.
• Importance of normalization??
  
  Better results with Euclidean distance
  DPD Worse results
  with Jeffrey divergence

22
Results 2D, Unweighted spectral data

• Euclidean
• Energy
• K-L
• Peak
• DPD
• JD

23
Results 2D, Weighted spectral data

• Euclidean
• Energy
• K-L
• Peak
• DPD
• JD

24
Conclusions II

• In the case of two-dimensional SOM better
  results are achieved by using non-weighted
  images.
• When the weighted images are used, the use of
  1D- SOM seems to be more reasonable.

25
Results histogram-trained 2D-SOM

• Euclidean
• Energy
• K-L
• Peak
• DPD
• JD

26
Connections between images and histograms
non-weighted
weighted
27
Past, Present Future

• Past What you have seen so far...
• Present Texture features in addition to color
  features
• Future Testing the effect of different metrics
  in ordering and fine-tuning
  phases (during the training of SOM)

28
Questions
29

30
(No Transcript)