Measurement of object structure from size-encoded images generated by optically-implemented Gabor filters

We use optical Fourier processing based on two dimensional (2D) Gabor filters to obtain size-encoded images which depict with 20nm sensitivity to size while preserving a 0.36μm spatial resolution, the spatial distribution of structural features within transparent objects. The size of the object feature measured at each pixel in the encoded image is determined by the optimal Gabor filter period, S(max), that maximizes the scattering signal from that location in the object. We show that S(max) (in μm) depends linearly on feature size (also in μm) over a size range from 0.11μm to 2μm. This linear response remains largely unchanged when the refractive index ratio is varied and can be predicted from numerical simulations of Gabor-filtered light scattering. Pixel histograms of the size-encoded images of isolated spheres and diatoms were used to generate highly resolved size distributions ("size spectra") exhibiting sharp peaks characterizing the known major structural features within the studied objects. Dynamic signal associated with changes in selected feature sizes within living cells is also demonstrated. Taken together, our data suggest that a label-free, direct and objective measurement of sample structure is enabled by the size-encoded images and associated pixel histograms generated from a calibrated optical processing microscope based on Gabor filtering.