Method for automatic separating useful and noise information in microscopic images particularly microscopic images of wafer surfaces

Claim 1

A method for automatic separation of useful and noise information in microscope images, particularly microscope images of wafer surfaces, for the purpose of automatic recognition of wafer identificatios characterized in reading a microscope image with a video camera; low pass filtering an image obtained by said reading (FIG. 1) to convert it into a low-contrast image (FIG. 2), standardizing the low pass filtered image by the existing gray tones according to the equation B3=(B1/B2-D)F to obtain a standardized image (FIG. 3) free of disturbing brightness fluctuations, where B1, B2 and B3 are mutually corresponding picture elements of the microscope, low contrast and standardized images (FIG. 1, FIG. 2 and, respectively, FIG. 3), and D and F are standardization constants defined by trial runs; producing a black/white image by evaluating the picture element information of the standardized image (FIG. 3) free of brightness fluctuations, using a permanently prescribed picture element brightness threshold; approximating both the edges of noise and useful components existing in the black/white image (FIG. 4) using polygon strokes with a tangential point identification method (FIG. 5) so that, repsective, tangential points for a saddle (S), a branching (F), a bottom point (B) or a top point (T) are identified when scanning the edges in four directions (0°, 45°, 90°, 135°), for which purpose the image content is converted into a data structure containing a digital image which can be plotted for test purpose (FIG. 6); eliminating the noise still present in the data structure (cf. FIG. 6) from the digital image which are eliminated by simple, predetermined dimensioning criteria so that the original identification structure free of noise elements can be accurately reconstructed (FIG. 7); and automatically identified.

Claim 2

A method for automatic separation of useful and noise information in microscope images, particularly microscope images of wafer surfaces, for the purpose of automatic recognition of wafer identificatios characterized in reading a microscope image with a video camera; low pass filtering an image obtained by said reading (FIG. 1) to convert it into a low-contrast image (FIG. 2), standardizing the low pass filtered image by the existing gray tones according to the equation B3=(B1/B2-D)F to obtain a standardized image (FIG. 3) free of disturbing brightness fluctuations, where B1, B2 and B3 are mutually corresponding picture elements of the microscope, low contrast and standardized images (FIG. 1, FIG. 2 and, respectively, FIG. 3), and D and F are standardization constants defined by trial runs; producing a black/white image by evaluating the picture element information of the standardized image (FIG. 3) free of brightness fluctuations, using a permanently prescribed picture element brightness threshold; approximating both the edges of noise and useful components existing in the black/white image (FIG. 4) using polygon strokes with a tangential point identification method (FIG. 5) so that, repsective, tangential points for a saddle (S), a branching (F), a bottom point (B) or a top point (T) are identified when scanning the edges in four directions (0°, 45°, 90°, 135°), for which purpose the image content is converted into a data structure containing a digital image which can be plotted for test purpose (FIG. 6); eliminating the noise still present in the data structure (cf. FIG. 6) from the digital image which are eliminated by simple, predetermined dimensioning criteria so that the original identification structure free of noise elements can be accurately reconstructed (FIG. 7); and automatically identified.