A real-time Non-Uniformity Correction(NUC) algorithm for a single-frame infrared image was proposed based on multi-thread optimization to correct the strip-type Fixed Pattern Noise(FPN) generated by non-uniformity response from an infrared camera. Firstly, the plus-relationship of the output value and the noise was determined by NUC algorithm based on total variation theory and a mathematical model was established to describe the strip noise. The image was split into a high-frequency component and a low-frequency component by filters, and the strip noise was fitted from the high-frequency component by the mathematical model. Finally, the real infrared image was restored by the plus-relationship obtained from the total variation theory. To improve the computing speed, the proposed algorithm was optimized with the multi-thread. The image quality evaluation of the proposed algorithm was carried out and its performance was compared with the Midway Infrared Equalization(MIRE) and the bilateral filter methods. Experimental results indicate that the proposed algorithm has reduced the response non-uniformity by 3% as compaired with original image according to national standard GB/T 17444-1998, and its efficiency is no difference with that of the MIRE. Moreover, system running time for an 14 bit infrared image with a resolution of 320×256 is 1.5 ms/frame, which meets the engineering requirements.