A high-level image sequence fusion algorithm for human detection

For detecting a human from background more clearly, we try to fuse two video image sequences captured from visible and thermal sensors respectively into one by a high-level image sequence fusion (HISF) algorithm. The HISF algorithm combined the Histogram of Oriented Gradient (HOG) descriptor and linear Support Vector Machine (SVM) aiming at high quality of human detection. Comparing with several multi-resolution image fusion algorithms, it is shown by the Detection Error Tradeoff (DET) curves that the proposed algorithm increases system robust performance and achieves an excellent classification rate.