Leukocyte image segmentation using feed forward neural networks with random weights

As we know, segmentation is an important countermeasure in the study of automated leukocyte image recognition. This paper proposes a novel method for leukocyte image segmentation, which is based on converting the segmentation to a classification issue. First, an effective classifier called feed forward neural network with random weights is employed to classify all the pixels in a leukocyte image. Then, according to the classification results, the regions of nucleus and cytoplasm are extracted, respectively, to achieve the segmentation. The experiments show that the proposed method is more effective compared with some existing approaches, and can segment the nucleus and cytoplasm well. Meanwhile, the advantage of the proposed method in leukocyte recognition is also reviewed and analyzed.