Ultra-wideband (UWB) microwave imaging is a very promising method for the early breast cancer detection physically based on the large contrast of electromagnetic parameters between the malignant tumor and the normal breast tissues. In this study, the tumor detection in a two dimensional breast model with hemi-elliptical configuration is carried out numerically. The influences from the skin, breast gland and the chest wall are involved in the study. The dispersion characteristics of the breast organisms are taken into account to approach the actual properties of the human breast. Results show that the tumor could be clearly recognized from the reconstructed images created by the confocal algorithm after the appropriate signal processing.