Multi-scale kernel local normalization for infrared image background suppression

Complex natural background(e.g. clouds and ground)suppression is a difficult problem for dim and small target detection in infrared image sequences. A dim and small target background suppression method based on wave atoms as a variant of two-dimensional (2D) wavelet packets, is proposed to solve the problem. It adopts kernel local normalization after wave atoms analysis, to suppress background details which contain edge, contour and texture, and enhance target information, and then modified coefficients are reconstructed using wave atoms inverse transform for suppression background. Experimental results demonstrate that, compared with wavelet transform (WT) and max-median (MMed) filter methods, the proposed method can suppress complex background in dim and small target images effectively. The improvement in signal-to-clutter ratio (ISCR) and background suppression factor (BSF) increase more than 3 times and 4 times, respectively.