Small target detection is a critical problem in the Infrared Search And Track (IRST) system.Although it has been studied for years, there are some difficulties remained due to the clutter environment such as the cloud edge and the horizontal line.In the homogeneous area such as sky, cloud-inner area and sea surface area, target can easily be detected, but in heterogeneous area which contains cloud edge, sky-sea line the target may be falsely detected.This paper proposes a novel method called kernel based local learning method to detect infrared small target in heavy clutter.First, each pixels estimation value is computed by using sliding window manner.Second, small targets are detected in the residual image which subtracts the estimation image from original image.Experimental results show our method achieves better performance.