Mosaic algorithm for large array infrared images sensor is studied. Aiming at problem of too much noise in IR image

Harris is adopted for noise inhibition and Harris algorithm is improved. Firstly

adaptive threshold is proposed according to characteristic of pixels response function value manual threshold of Harris detection is improved to adaptive threshold. On the basis of similar measure normalized cross correlation（ NCC） coarse matching

random sampling consensus（ RANSAC） algorithm is adopted which based on constraints. According to correctly matched corner point between images each constructive quadrilateral should meet geometic relationship of four angles are equal. Experimental result shows the algorithm avoid effect of noise

can make image mosaic has adaptability

it has good mosaic effect for large array infrared image.