TY  - CONF
EP  - 3080
VL  - 2015-N
A1  - Khan, J.
A1  - Malik, A.S.
A1  - Kamel, N.
A1  - Dass, S.C.
A1  - Affandi, A.M.
UR  - https://www.scopus.com/inward/record.uri?eid=2-s2.0-84953284847&doi=10.1109%2fEMBC.2015.7319042&partnerID=40&md5=c60d97b51966565fb4b2719e2a77e98a
PB  - Institute of Electrical and Electronics Engineers Inc.
SN  - 1557170X
Y1  - 2015///
TI  - Segmentation of acne lesion using fuzzy C-means technique with intelligent selection of the desired cluster
ID  - scholars5629
SP  - 3077
KW  - acne vulgaris; algorithm; automated pattern recognition; cluster analysis; color; dermatology; fuzzy logic; human; image processing; light; pathology; procedures; reproducibility; sensitivity and specificity; skin
KW  -  Acne Vulgaris; Algorithms; Cluster Analysis; Color; Dermatology; Fuzzy Logic; Humans; Image Processing
KW  -  Computer-Assisted; Light; Pattern Recognition
KW  -  Automated; Reproducibility of Results; Sensitivity and Specificity; Skin
N1  - cited By 18; Conference of 37th Annual International Conference of the IEEE Engineering in Medicine and Biology Society, EMBC 2015 ; Conference Date: 25 August 2015 Through 29 August 2015; Conference Code:116805
N2  - Segmentation is the basic and important step for digital image analysis and understanding. Segmentation of acne lesions in the visual spectrum of light is very challenging due to factors such as varying skin tones due to ethnicity, camera calibration and the lighting conditions. In this approach the color image is transformed into various color spaces. The image is decomposed into the specified number of homogeneous regions based on the similarity of color using fuzzy C-means clustering technique. Features are extracted for each cluster and average values of these features are calculated. A new objective function is defined that selects the cluster holding the lesion pixels based on the average value of cluster features. In this study segmentation results are generated in four color spaces (RGB, rgb, YIQ, I1I2I3) and two individual color components (I3, Q). The number of clusters is varied from 2 to 6. The experiment was carried out on fifty images of acne patients. The performance of the proposed technique is measured in terms of the three mostly used metrics; sensitivity, specificity, and accuracy. Best results were obtained for Q and I3 color components of YIQ and I1I2I3 color spaces with the number of clusters equal to three. These color components show robustness against non-uniform illumination and maximize the gap between the lesion and skin color. © 2015 IEEE.
AV  - none
ER  -