A PCNN improved with fisher criterion for infrared human image segmentation

Abstract

Infrared human image is not enough contrasted, and gray overlap between background and human targets. A PCNN improved with Fisher criterion image method is proposed for these problems. The fisher criterion-based segmentation method achieve good segmentation in low image contrast. PCNN segmentation method is suitable for image with gray overlap between object and background. Combined these two method solves these two problems. Setting up dynamic threshold weight factor solves under-segmentation and over-segmentation problem of human body infrared image. Results show our method outperforms classical methods in terms of segmentation effect and rate.