High-quantization-bit sea surface infrared image segmentation method based on self-organizing mapping

The invention provides a high-quantization-bit sea surface infrared image segmentation method based on self-organizing mapping. The method comprises the following steps: calculating a preprocessed image according to an input infrared image; calculating a pixel feature matrix according to the preprocessed image; according to the pixel feature matrix, self-organizing mapping network calculation learning is carried out, and a classification result is obtained through calculation; calculating a segmented connected graph according to a classification result; and calculating and outputting a segmented image according to the segmented connected graph. The invention provides a brand-new pixel feature combining a pixel gray value, a statistical value and a difference absolute value, and accurate segmentation of a high-quantization-bit sea surface infrared image can be efficiently realized; a self-organizing mapping unsupervised learning method is adopted, extra labels are not needed, automatic clustering can be achiev