Automatic Classification of White Blood Cells(Feature Extraction Part 2)

The present paperdeals with the following problems for the automatic white blood cell classification. (1)A technique for extracting a nucleus region of the white blood cell is described. It can be accomplishedby using the histogram of the directional difference. From this histograrm，an optimum threshold value for extracing the nucleus region is obtained. (2)The texture parmeters which express effectively the difference of the texture between white blood cells are proposed. That is，the shape of the chromatin in the nucleus and the amount of the granules are proposed to discriminate the differene between monocytes and large lymphocytes and the difference between eosinophils and basophils， respectively