Skip to main page content
U.S. flag

An official website of the United States government

Dot gov

The .gov means it’s official.
Federal government websites often end in .gov or .mil. Before sharing sensitive information, make sure you’re on a federal government site.

Https

The site is secure.
The https:// ensures that you are connecting to the official website and that any information you provide is encrypted and transmitted securely.

Access keys NCBI Homepage MyNCBI Homepage Main Content Main Navigation
. 2014;24(7-8):1759-1770.
doi: 10.1007/s00521-013-1416-9. Epub 2013 Apr 27.

An improved SOM algorithm and its application to color feature extraction

Li-Ping Chen  1 , Yi-Guang Liu  2 , Zeng-Xi Huang  2 , Yong-Tao Shi  2
Affiliations

An improved SOM algorithm and its application to color feature extraction

Li-Ping Chen et al. Neural Comput Appl. 2014.

Abstract

Reducing the redundancy of dominant color features in an image and meanwhile preserving the diversity and quality of extracted colors is of importance in many applications such as image analysis and compression. This paper presents an improved self-organization map (SOM) algorithm namely MFD-SOM and its application to color feature extraction from images. Different from the winner-take-all competitive principle held by conventional SOM algorithms, MFD-SOM prevents, to a certain degree, features of non-principal components in the training data from being weakened or lost in the learning process, which is conductive to preserving the diversity of extracted features. Besides, MFD-SOM adopts a new way to update weight vectors of neurons, which helps to reduce the redundancy in features extracted from the principal components. In addition, we apply a linear neighborhood function in the proposed algorithm aiming to improve its performance on color feature extraction. Experimental results of feature extraction on artificial datasets and benchmark image datasets demonstrate the characteristics of the MFD-SOM algorithm.

Keywords: Color feature extraction; Competitive mechanism; Non-principal component; Self-organizing map.

PubMed Disclaimer

Figures

Fig. 1
Fig. 1
Demonstrations of two kinds of competitions. a Virtual training samples. b A possible result trained by the conventional SOM. c A possible intermediate result trained by MFD-SOM. d A possible final result trained by MFD-SOM (color figure online)
Fig. 2
Fig. 2
Two ways of updating weight vectors of neurons. a Traditional way. b Proposed way (color figure online)
Fig. 3
Fig. 3
Framework of the MFD-SOM algorithm (color figure online)
Fig. 4
Fig. 4
Two artificial datasets and features extracted from them by both algorithms. a Two artificial datasets. b Features extracted by the conventional SOM. c Features extracted by our algorithm
Fig. 5
Fig. 5
Training images. a River (140055.jpg). b Flower (118_0081.jpg). c Bird (049_0097.jpg). d Cloth (257_0178.jpg). e Girl (253_0354.jpg). f Penguin (158_0135.jpg)
Fig. 6
Fig. 6
Demonstrations of extracted color features. Thumbnails of training images (in the left column), color features maps achieved by the conventional SOM algorithm (in the middle column) and achieved by our algorithm (in the right column)
Fig. 7
Fig. 7
3D visualizations of colors achieved by both algorithms from the “river” image. a Colors extracted by the conventional SOM. b Colors achieved by the proposed SOM (color figure online)
Fig. 8
Fig. 8
Reconstructed images with reference to the “river” image. a Original image. b Using colors achieved by the conventional SOM after 200 training epochs. c Using colors achieved by MFD-SOM after 20 training epochs. d Using colors achieved by the conventional SOM after 20 training epochs (color figure online)
Fig. 9
Fig. 9
Sets of color feature maps of the “river” image achieved by both algorithms. a Maps achieved by the conventional SOM. b Maps achieved by the MFD-SOM (color figure online)
Fig. 10
Fig. 10
Demonstration of colors extracted from image dataset. a 40,000 color samples. b Color feature map purified by the conventional SOM for (a). c Color feature map purified by our algorithm for (a) (color figure online)
See this image and copyright information in PMC

References

    1. Kohonen T. The self-organizing map. Neurocomputing. 1998;21(1998):1–6. doi: 10.1016/S0925-2312(98)00030-7. - DOI
    1. Kohonen T (2001) Self-organizing maps, 3rd edn. Springer Ser Inf Sci 30:501
    1. Yin H (2008) The self-organizing maps: background, theories, extensions and applications. In: Computational intelligence: a compendium, vol 115. Springer, Heidelberg, pp 715–762
    1. Moschou V, Ververidis D, Kotropoulos C. Assessment of self-organizing map variants for clustering with application to redistribution of emotional speech patterns. Neurocomputing. 2007;71(1–3):147–156. doi: 10.1016/j.neucom.2007.08.013. - DOI
    1. Ong SH, Yeo NC, Lee KH, Venkatesh YV, Cao DM. Segmentation of color images using a two-stage self-organizing network. Image Vis Comput. 2002;20(4):279–289. doi: 10.1016/S0262-8856(02)00021-5. - DOI

LinkOut - more resources