Quadratic blur kernels for latent image formation modeling
Anatoly Burov · Proceedings of SPIE, the International Society for Optical Engineering/Proceedings of SPIE · 2010
A bilinear photoresist model is accurate, fast, and potentially reversible. Similar to other image-processing style (blur kernel) models, this model represents a transformation of an aerial image into a latent image. The key difference is the explicit recognition of the non-linearity of the process while retaining common signal processing architecture. By applying a Volterra series expansion to the reaction-diffusion functional, a high-accuracy representation of the process is obtained. Several methods for identifying the double-impulse response of the quadratic term of the series are discussed. Characterization is carried out based on the bi-harmonic signal sampling method of the Bilinear Transfer Function, the Fourier transform of the double-impulse spread function. Several photoresist systems are characterized, and strong quadratic behavior is observed for many. The resulting estimated BTF are presented, and their differences are discussed.