Automatic Recognition And Counting Of Suspension Cells Based On Regional Histogram

It is needed to count their live cells after suspension cells are cultured. The cells arenormally counted by a method called "cell-count boards". It means that human's eyesare employed to count some live cells one by one. Such method costs the operatormuch time and energy, and he/she must be proficient in operating. In order to solvethese problems of counting cells by manual operation, we try to recognize and count thelive cells automatically by some image-analyzing techniques. In this paper, cell imageswere taken under the microscope with magnification 100. And these images can beclassified into two kinds: low contrast with few cells, high contrast with many cells intheir corresponding images. The first kind of image was processed in the thesis bymeans of background fitting, but such method is useless for the second kind of image.So a different method are attempted to process the second one. Image segmentationshould be usually done before cell recognition. Several segmentation methods are tried,but the results are not good enough. In this paper, our image is successfully segmentedbased on a local histogram. Then some parameters such as area, circumference andform factor of every cell can be obtained by tracking a segmented image by means ofchain code. With these parameters, all cells in a cell image can be tracked and countedeffectively and efficiently.