Image threshold segmentation is very important in image analysis and recognition

in view of the problem of maximal variance between-class method—Otsu method that the background and target information being erroneously segmented when object size is relative small in complex background or the object and background has small gray level difference

an improved maximal variance between-class method is put forward.The method is to make a normalized projection for the preprocessing infrared warship image

the image with a certain threshold is cut down

Otsu method is used to segment the image.A comparison among the Otsu method

the recursive Otsu method and this method is carried out

the experimetal results show that the method has a good adaptive and best segmentation performance.