Performance of the PCAN+NN in the presence of optical noise

Abstract

Optical noise is present in any stage of an imaging system. For instance, images obtained from laser-illuminated objects suffer from speckles. Images of objects beneath a highly-scattering medium also have low signal-to-noise ratios and are therefore difficult to distinguish. In this paper we assess the recognition rate of a Principal Components Analyzing Network plus Neural Network (PCAN+NN) trained to recognize images of metaphase spreads, cells and empty slides if these images are superimposed with optical noise.