Infrared small target detection method based on frequency domain clutter suppression and spatial feature extraction

With the continuous development of infrared technology, application imagery is increasingly widespread. Nonetheless, suffers from low contrast and high noise characteristics, making it challenging to detect recognize small targets. Frequently, existing deep learning-based target detection methods solely employ spatial information, neglecting frequency domain thereby creating susceptibility background clutter interference accuracy in complex scenes. Here, a method proposed that combines suppression feature extraction Firstly, image transformed into using block discrete cosine transform (DCT). Secondly, information filtered an attention mechanism adaptively enhance targets suppress clutter. Finally, compressed map back inputted lightweight neural network for fusion obtain results. This approach enhances accuracy, reduces computational complexity storage space, practical deployable. Experimental outcomes demonstrate strong capabilities on public datasets, significant advantages over traditional learning methods.