Fig. 1. Flowchart of the improved template matching algorithm

Fig. 2. Two typical infrared cross targets：(a) nested crosses in a cross; (b) nested dots in a cross

Fig. 3. Schematic diagram of positioning deviation due to cross deflection：(a) small deflected cross with uncorrected kernel; (b) large deflected cross with uncorrected kernel; (c) large deflected cross with corrected kernel

Fig. 4. Pyramid template matching

Fig. 5. Typical infrared test pictures：(a) normal cross image; (b) noise-enhanced image; (c) blur-enhanced image; (d) reduced cross image; (e) enlarged incomplete image; (f) bright block background image

Fig. 6. The results of the algorithm in this paper

Fig. 7. Comparison of the filtering effect of each algorithm