Glomerulus extraction is an important step for automatic analysis of the kidney diseases in the computer-aided diagnosis system. A method based on searching the best fitting curve is proposed based on the characteristics of the renal biopsy images in microscope. This method can solve the problem of the large defect of the enhanced boundary, which lead to unsuccessful extraction. Firstly, a parametric equation is constructed based on the cubic spline interpolation function to draw the closed fitting curve. Secondly, the different scale binary images can be obtained by adjusting the parameters of the LOG filter. Finally, after labeling to remove the noises and thinning, a genetic algorithm is used to search the best flitting curve for glomerulus boundary. Experimental results indicate the correctness and effectiveness of this method.