To overcome the problems of insensitiveness to noise and low detection rate in original Harris algorithm,a novel corner detection algorithm based on bilateral kernel function is presented.Firstly bilateral filter is used to replace the Gauss lowpass filter,which improves the robustness of Harris algorithm.Then the multi-scale decomposition is applied to produce segmentation threshold of the true corner and the false corner.Experimental results show that the proposed algorithm can precisely detect comers in images.