| CPC G06T 5/70 (2024.01) [G06T 5/20 (2013.01); G06T 7/0006 (2013.01); G06T 2207/10061 (2013.01); G06T 2207/20081 (2013.01); G06T 2207/20084 (2013.01); G06T 2207/30148 (2013.01)] | 9 Claims |
|
1. A method of denoising electron image measurement waveforms of a pattern, the method comprising:
scanning an electron beam across a pattern of interest on a substrate;
detecting a single scan line waveform of the scanned electron beam;
obtaining a model one-dimensional (1D) scan line waveform from the detected scan line waveform;
augmenting the (1D) model scan line waveform by performing at least one of squeezing, widening, skewing and shifting the model scan line waveform to obtain clean (1D) augmented model data and adding noise to the at least one of a squeezed, widened, skewed and shifted versions of the (1D) model scan line waveform to obtain noisy augmented data;
applying a deep learning neural network process to the noisy (1D) augmented data;
comparing the output of the deep neural network training process to the clean (1D) augmented model data to obtain a minimal square error; and
iteratively updating parameters of the deep neural network training process by backpropagating the obtained minimal square error obtained into the deep neural network training process to create a noise discrimination function.
|