In this work we have used a number of texture descriptors to characterize the acrosome state of boar sperm cells, which is a key factor in semen quality control applications. Laws masks, Legendre and Zernike moments, Haralick features extracted from the original image and from the coefficients of the Discrete Wavelet Transform, and descriptors based on interest points using the Speeded-Up Robust Features (SURF) method have been evaluated. Classification using kNN show that the best results were obtained by SURF, with an overall hit rate of 94.88% and, what is more important, a higher hit rate in the damaged (96.86%) than in the intact class (92.89%). These results make this descriptor very attractive for the veterinary community.
|Title of host publication||Computational Vision and Medical Image Processing, Proceedings of VipIMAGE 2011 - 3rd ECCOMAS Thematic Conference on Computational Vision and Medical Image Processing|
|Number of pages||5|
|Publication status||Published - 13 Feb 2012|
|Event||3rd ECCOMAS Thematic Conference on Computational Vision and Medical Image Processing, VipIMAGE 2011 - Olhao, Algarve, United Kingdom|
Duration: 12 Oct 2011 → 14 Oct 2011
|Conference||3rd ECCOMAS Thematic Conference on Computational Vision and Medical Image Processing, VipIMAGE 2011|
|Period||12/10/11 → 14/10/11|