Infrared Small Target Detection Based on Singularity Analysis and Constrained Random Walker

Effective infrared small target detection is still challenging due to sizes and the clutter in background. Unfortunately, many advanced methods do not perform well preserving detecting multi-scale objects complex scenes. We propose an targets method suppress background adapt with different sizes. Based on singular value analysis facet model, we a multi-derivative descriptor enhance various dual derivative channels. In first-order channel, design four kernels directions preserve isotropic block clutter. second-order use kernel center pixels of band order sizes, constrained random walker technique, including adaptive matching algorithm extract local regions each candidate adaptively based constraint size shape. The experimental results demonstrate that proposed can accurately detect scenes, resulting better performance than state-of-the-art methods.