Homomorphic vs. multiplicative lateral inhibition models for image enhancement

Homomorphic and multiplicative lateral inhibition models of adaptive image enhancement filters are examined theoretically and empirically. A Volterra-Wiener series expansion of both model identifies some of the mechanisms responsible for their inherently adaptive behaviour and resulting from their nonlinear nature. The empirical study compares their image enhancement in terms of both local DC gain (local luminance mapping) and local contrast enhancement (local perceived contrast gain). While both operators exhibit a degree of locally adaptive behaviour we conclude the richer nonlinear structure of the multiplicative inhibition model allows for more dramatic adaptive enhancement capability.<>