With the use of computer image analysis (IA) and manual counting, the immunohistochemistry (IHC) results on PS2, ER and c-erB-2 of breast cancer which represent positive expression of cytoplasm, cell nucleus and cytomembrane are analyzed and compared. Then the advantages, disadvantages and repetition of the two methods are discussed and the correlation between image analysis and manual counting is estimated. The results of this study indicate that IA has better repetition in estimation of positive expression of cytoplasm, cell nucleus and cytomembrane, whereas manual counting has better repetition only in the positive expression of cell nucleus and cytomembrane, and the two methods have positive correlation in estimation of IHC results. In addition, some problems in image processing are discussed in this paper.