Feedback and Isolation
The PCNN can be a very powerful front-end processor for an image recognition system. This is not surprising since the PCNN is based on the biological version of a pre-processor. The PCNN has the ability to extract edge information, texture information, an
- PDF / 1,119,000 Bytes
- 19 Pages / 439.37 x 666.142 pts Page_size
- 54 Downloads / 174 Views
Feedback and Isolation
6.1 A Feedback PCNN The PCNN can be a very powerful front-end processor for an image recognition system. This is not surprising since the PCNN is based on the biological version of a pre-processor. The PCNN has the ability to extract edge information, texture information, and to segment the image. This type of information is extremely useful for image recognition engines. The PCNN also has the advantage of being very generic. Very few changes (if any) to the PCNN are required to operate on different types of data. This is an advantage over previous image segmentation algorithms, which generally require information about the target before they are effective. There are three major mechanisms inherent in the PCNN. The first mechanism is a dynamic neural threshold. The threshold, here denoted by Θ, of each neuron significantly increases when the neuron fires, then the threshold level decays. When the threshold falls below the respective neuron’s potential, the neuron again fires, which raises the threshold, Θ. This behaviour continues which creates a pulse stream for each neuron. The second mechanism is caused by the local interconnections between the neurons. Neurons encourage their neighbours to fire only when they fire. Thus, if a group of neurons is close to firing, one neuron can trigger the entire group. Thus, similar segments of the image fire in unison. This creates the segmenting ability of the PCNN. The edges have different neighbouring activity than the interior of the object. Thus, the edges will still fire in unison, but will do so at different times than the interior segments. Thus, this algorithm isolates the edges. The third mechanism occurs after several iterations. The groupings tend to break in time. This “break-up” or de-synchronisation is dependent on the texture within a segment. This is caused by minor differences that eventually propagate (in time) to alter the neural potentials. Thus, texture information becomes available. The Feedback PCNN (FPCNN) sends the output information in an inhibitory fashion back to the input in a similar manner to the rat’s olfactory system. The outputs are collected as a weighted time average, A, in a fashion similar to the computation T. Lindblad and J. M. Kinser, Image Processing Using Pulse-Coupled Neural Networks, Biological and Medical Physics, Biomedical Engineering, DOI: 10.1007/978-3-642-36877-6_6, © Springer-Verlag Berlin Heidelberg 2013
107
108
6 Feedback and Isolation
of Θ except for the constant V , Ai [n] = e−α A δn Ai j [n − 1] + V A Yi j [n],
(6.1)
where V A is much lower than V and in this case V A = 1. The input is then modified by, Si j [n − 1] . (6.2) Si j [n] = Ai j [n − 1] The FPCNN iterates the PCNN equations with Eqs. (6.1) and (6.2) inserted at the end of each iteration. Two simple problems are shown to demonstrate the performance of the FPCNN. The first problem used a simple square as the input image. Figure 6.1 displays both the input stimulus S and the output Y for a few iterations until the input stabilised.
Data Loading...
