|
|||||||
Malignancy Texture Classification in Digital Mammograms based on Chebyshev Moments and Logpolar Transformation Vibha S. Vyas and Priti P.Rege College of Engineering Pune Shivajinagar, Pune-411005, India Phone: +91-20-25507165, Fax+91-20-25507299 Abstract: In this paper, an algorithm is proposed for the classification of malignancy in mammogram. The types of malignancy can be classified as circumscribed, speculated, architectural distortion, miscellaneous, asymmetry and calcification. Chebyshev moments (CM), being discrete, are capable to combat the discritisation errors as compared to Zernike Moments (ZM), which are continuous moments. It is also orthogonal and can very well be represented in terms of Geometric moments (GM) to achieve rotation scale and translation (RST) invariant textural analysis. A novel approach to achieve RST invariance using combination of CM and Logpolar coordinate system is proposed. As Logpolar transformation (LPT) is well known for rotation and scale invariance, we need not express CM in terms of GM. The translation invariance is archived by shifting the LPT image to centroid before CM calculation. The average classification rate is compared with Circular Mellin Feature extractors. The images are taken from the Mammographic Image Analysis Society’s (MIAS) Mini-Mammographic database. Keywords: Chebyshev moments, Logpolar transformation, mammograms, rotation, scale and translation invariance (RST invariance).
(
Biography:
BibTex: @ARTICLE{P1160741002, AUTHOR = {Vibha S. Vyas and Priti P.Rege}, TITLE = {Chebyshev Moments Based Texture classification of Malignant Masses in Digital Mammograms},
JOURNAL = {ICGST International Journal on Bioinformatics and Medical
Engineering,
BIME},
YEAR = {2007}, VOLUME = {7}, ISSUE ={1}, PAGES={29--35} }
( |
|||||||
|