A method for designing a filter to image a feature on a surface, comprising: acquiring an image of said feature, with said image of feature comprising information from multiple points of said feature; generating a structural model of said feature by extracting predetermined properties of said feature from said image of feature; computing a scattering model for said feature from said structural model of said feature, with said scattering model for feature having information on scattered electromagnetic field from feature propagating in a plurality of scattering angles, wherein said scattered electromagnetic field from feature is generated by scattering of an electromagnetic radiation by said feature; acquiring an image of said surface, with said image of surface comprising information from multiple points of said surface; generating a structural model of said surface by extracting predetermined properties of said surface from said image of surface; computing a scattering model for said surface from said structural model of said surface, with said scattering model for surface having information on scattered electromagnetic field from surface propagating in a plurality of scattering angles, wherein said scattered electromagnetic field from surface is generated by scattering of an electromagnetic radiation by said surface; and computing said filter by combining said scattering model for feature and said scattering model for surface to achieve a predetermined filter performance metric, whereby said filter is designed to modulate scattered electromagnetic field from said feature and scattered electromagnetic field from said surface to image a feature on said surface. A system and method for recognizing a feature, comprising: acquiring an image of said feature using an imaging module, with said image of feature comprising information from multiple points of said feature; computing a feature spread function from scattering model of a previously known feature and transfer function of said imaging module, wherein said feature spread function represents a model of an image of said previously known feature; and comparing said image of feature with said feature spread function by computing a match metric between said image of feature and said feature spread function, whereby said match metric determines if said feature is similar to said previously known feature.