The non-effective pixels exist in the un-cooled infrared focal plane array (UFPA) inevitably, which has a bad effect on the UFPA imaging. Based on analyzing some kinds of algorithm on the non-effective pixels detection of un-cooled infrared focal plane array, a new method of detection and real-time compensation is presented in this article to solve this problem. First of all, the best threshold of the non-effective pixels detection is searched according to the pixels′ response characteristics and the locations of the non-effective pixels are marked as well. When realizing on hardware, for an M×N UFPA, at any sampling time, a shift-register is used to store M data which is before the current sample data and the shift-register′s output is updated with the data which is at the same column while the upper row. Meanwhile, a common register is used to store the data which is at the same row while the previous column. Then, the interpolation theory is introduced to compensate the non-effective pixel. The approach proposed solves the difficulty of the non-effective pixels detection and compensation efficiently. For a 320×240 UFPA, the algorithm has been realized on an infrared image processing system based on FPGA and the effect of non-effective pixel on the imaging system is eliminated successfully.