Studies on segmentation and recognition marrow cells image

Aiming at the characteristics of the marrow cell image, proposed a new method to calculate local regularity and got a appropriate threshold using multi-fractal spectrum so that accurate edges of different kinds of cells can be detected. Adopted two neural networks: using principal component analysis based neural network (GHA) to acquire three principal components and using simulated annealing (SA) and BP network to class and recognize the marrow cells.