Abstract: In order to correct the stripe nonuniformity for infrared images captured by infrared linear arrays and uncooled staring infrared focal plane arrays, a novel stripe nonuniformity correction algorithm based on steering kernel for single infrared image was proposed. Firstly, the principal directions and window sizes of steering kernels were determined by the gradient characteristics of infrared images. Then, the estimations of each column pixel were treated as expectations in least square fitting, which was in the constraint of vertical stripe models. The correction parameters in the iteration were obtained, and the image corrected was accepted with a certain probability. The iteration of nonuniformity correction should not be ended until the error function is below the threshold preset. Experimental results indicate that the proposed algorithm has stable and fast convergence. Compared with other three algorithms, the proposed algorithm has the advantage of reducing the stripe nonuniformity and preserving much more edge information.