Abstract: Aiming at the severe problems of system performance degradation that was caused by the blind pixels and the nonuniformity which exists in Infrared Focal Plane Array (IRFPA), multivariate normal distribution sequence noise model of IRFPA was set up firstly, and blind-pixel was regarded as abnormity pixel for the unconformiting mold statistical distribution characteristic and it was dissociating the multivariate normal distribution ellipsoid. Then the principal component was applied to sequence pattern, and the statistical distance and bisectrix spatial angle are regard as statistical criterion of abnormity pixels detection. Finally, to test and verify the performance of the method on taken full advantage of thermal infrared imager to do sequential multiple frames of noise data collection. The algorithm was applied to actual blind pixel detection of uncooled IRFPA and the validity of the algorithm was proved by the experimental result.