Study on real-time registration in dual spectrum low level light night vision technique

Abstract

In low level light (LLL) color night vision technology, dual spectrum images with respective special information were acquired, and target identification probability would be effectively improved through dual spectrum image fusion. Image registration is one of the key technologies during this process. Current dual spectrum image registration methods mainly include dual imaging channel common optical axis scheme and image characteristic pixel searching scheme. In dual imaging channel common optical axis scheme, additional prismatic optical components should be used, and large amount of radiative energy was wasted. In image characteristic pixel searching scheme, complicated arithmetic made it difficult for its real time realization. In this paper, dual channel dual spectrum LLL color night vision system structure feature and dual spectrum image characteristics was studied, dual spectrum image gray scale symbiotic matrix 2-dimensional histogram was analysed, and a real time image registration method including electronic digital shifting, pixel extension and extraction was put forward. By the analysis of spatial gray-scale relativity of fusion image, registration precision is quantitatively expressed. Emulation experiments indicate that this arithmetic is fast and exact for our dual channel dual spectrum image registration. This method was realized on dual spectrum LLL color night vision experimental apparatus based on Texas Instruments digital video processing device DM642.