The invention discloses a segmentation method for overlapping cells in a cervical smear image. The method comprises the following steps that firstly, an image background is removed, a frontground object, namely a
cell aggregate formed by the overlapping cells, is extracted; secondly,
cell nucleuses are segmented; thirdly, color partition is carried out on the
cell aggregate by using the average value shifting method, and sub-blocks are formed by segmenting by using the
watershed algorithm; fourthly, all of combination
modes of the sub-blocks are searched, filtering is carried out according to a screening condition, the combinations which are in accordance with the condition are called possible solutions; fifthly, energy functionals of the possible solutions which are obtained in the fourth step are calculated, and a
minimum function value is selected as an optimal solution. The sub-blocks are regarded as units to select the optimal possible solution, and the known space is shrunk; an illumination model of a transmitting object overlapping portion is brought up, the theoretical foundation for discrimination of the overlapping sub-blocks is provided; the designed
energy functional is integrated with multiple characteristics of the cells, and therefore the selected optimal solution has the reasonability and the effectiveness; a brought up
processing frame can be used as a
common framework for segmenting the transmitting object overlapping portion.