Abstract: Based on the center-surround mechanism of biological vision, a method which uses an improved LoG (Laplacian of Gaussian) operator to detect light blobs is proposed. First of all, the imaging model of light blob targets was established. The properties of LoG operators and their response to image were analyzed and a normalized LoG operator was proposed. The normalized LoG operators can be used to detect light blobs according to theoretical deduction. Simulation and test show that the proposed operator is effective in detecting size-known and size-unknown blobs in an image. The proposed operator can also be used to track single target and rigid-body in image sequence if combined with Kalman filter or particle filter. This method is proven to be robust and effective.