PURPOSE: To apply pre-processing to easily extract a main characteristic component of a multi-variable data by applying convolution calculation to a multi- variable data with a prescribed discrete coefficient.
CONSTITUTION: An initial value setting section 1 sets a constant (standard deviation σ, lx, ly of the filter in x and y directions, x0, y0 normalizing them) required to decide the characteristic of a Laplacian Gaussian(LG) filter and sends the result to a calculation section 2. The calculation section 2 uses the constant from the setting section 1 to decide the characteristic and to give weighting to a 2-dimension G(x, y) function for smoothing and taking Laplacian. A discrete coefficient Wij of an equation obtained is fed to, an arithmetic section 3, which applies convolution calculation between the coefficient and the input data value f(m, n), an output data h(m, n) is outputted and a normalizing processing section 4 compares the data with a value normalized based on the standard variance of the data with a threshold level and the processing converting the characteristic extraction easily is taken.
TOKUDA KIYOHITO
SATO TAKURO
FUKAZAWA ATSUSHI
