ABSTRACT

In the 1990s, a new process of analyzing images emerged which modeled pulsing activity of neurons from the mammalian visual cortex. In an image processing application pulse patterns of the neurons were used to extract shapes, textures, motion, and a variety of other features in the image. This chapter reviews the theory of the pulse image generators and presents a few of the applications. The pulse-coupled neural network (PCNN) is a digital model based upon the visual cortex. When applied to images, each pixel is assigned to a neuron, and each neuron is connected to local neighbors. There have been several models of the mammalian visual cortex starting with the Hodgkin-Huxley model of the 1950s and the Fitshugh-Nagumo model in the early 1960s. These models described neural behavior as leaky integrators and mathematically through coupled differential equations that linked neural potentials. One of the main drawbacks the PCNN is the interference of expanding autowaves.