An adaptive algorithm based on quasi-Gaussian model is proposed for locating fixed noise in microscope image. Firstly, by considering temporal correlation of microscope image sequence, mean and variance of each color component for pixels at the same location in image sequence are computed. Secondly, the maximizing inter-class variance (OTSU) method is used to obtain three adaptive mean thresholds of color components. Meanwhile, three adaptive variance thresholds are acquired by averaging minimum variance and maximum variance. Then, the thresholds are utilized to check each pixel in the image sequence to detect the candidates of fixed noise. After several iterations to eliminate false candidates, fixed noises are finally located. Experimental results show that the proposed algorithm can stably detect fixed noise in microscope image sequences even though the background and object luminance change a lot.