Adaptive pixel targeting and filtering with adjacency variations in second-order derivatives of pixel values for SEM images

Abstract

The proposed method utilizes histograms of second-order derivatives of the values of the pixels in an SEM image. The histograms of second-order values can express the variation in pixel values due to the effect of noise. The histograms are used as the basis for a technique to target pixels for filtering. To control the number of pixels which are targeted so as to minimize blurring of edges, a curve-fitted profile is imposed on the histograms in order to select pixels based on the differences between their second-order derivatives and those of their neighbours. The proposed method is found to be more effective at higher levels of additive noise.