Inspection method and apparatus for the inspection of either random or repeating patterns

• US 5,537,669 A
• Filed: 09/30/1993
• Issued: 07/16/1996
• Est. Priority Date: 09/30/1993
• Status: Expired due to Term

• Current Assignee: KLA Instruments, Corporation
• Original Assignee: KLA Instruments, Corporation
• Inventors: Tsai, Bin-Ming B., Lin, Jason Z., Evans, David M. W.
• Primary Examiner(s): Boudreau, Leo
• Assistant Examiner(s): Tran, Phuoc
• Application Number:
  US08/129,341
• Time in Patent Office: 1,020 Days
• Field of Search: 382/43, 382/42, 382/48, 382/8, 382/280, 382/278, 382/282, 382/294, 382/270, 382/275, 382/141, 382/145, 382/149, 382/151
• US Class Current: 382/141
• CPC Class Codes: G03F 7/7065 Defects, e.g. optical inspection of patterned layer for defects
• CPC Class Codes: G06F 18/00 Pattern recognition
• CPC Class Codes: G06T 2207/20056 Discrete and fast Fourier transform, [DFT, FFT]
• CPC Class Codes: G06T 2207/30148 Semiconductor; IC; Wafer
• CPC Class Codes: G06T 7/0004 Industrial image inspection
• CPC Class Codes: G06T 7/37 using transform domain methods
• CPC Class Codes: G06V 10/431 Frequency domain transformation; Autocorrelation
• CPC Class Codes: G06V 10/751 Comparing pixel values or logical combinations thereof, or feature values having positional relevance, e.g. template matching

Abstract

The present invention is a hybrid technique for finding defects on digitized device images using a combination of spatial domain and frequency domain techniques. The two dimensional spectra of two images are found using Fourier like transforms. Any strong harmonics in the spectra are removed, using the same spectral filter on both spectra. The images are then aligned, transformed back to the spatial domain, and subtracted. The resulting spectrally-filtered difference image is thresholded and analyzed for defects. Use of the hybrid technique of the present invention to process digitized images results in the highest-performance and most flexible defect detection system. It is the best performer on both array and random devices, and it can cope with problems such as shading variations and the dark-bright problem that no other technique can address. The hybrid technique of the present invention also uses frequency domain techniques to align the images with fewer errors than spatial domain techniques of similar or lesser complexity. Further, the relative offsets of the pairs of images are determined by frequency domain techniques--and this method may be the most accurate and the least expensive.

Claims

1. In an inspection system for detecting defects in a device having either random, repeating or mixed patterns thereon, said inspection system includes first and second fields of view each positioned along a different optic axis, the first field of view creates a first spatial domain pixel map of a portion of a first device on a first substrate;

and said second field of view provides a second spatial domain pixel map against which said first spatial domain pixel map is to be checked;

said second spatial domain pixel map is developed from one of a pre-stored pixel map of a desired pattern, a scaled drawing of a desired pattern, a repeating pattern within said first device in said first field of view, a pattern within a second device on said first substrate as in said first field of view, and a pattern on a second substrate;

a method for detecting defects in said first device in said first field of view comprising the steps of;

(a) transforming said first and second spatial domain pixel maps from said first and second fields of view using a two dimensional Fourier like transform to convert said first and second spatial domain pixel maps from the spatial domain to first and second frequency domain pixel maps, respectively;

(b) identifying those frequencies in both dimensions having amplitude maxima in the frequency domain of said first and second frequency domain pixel maps with any repeating pattern of each of said first and second frequency domain pixel maps from said first and second fields of view occurring at those frequencies in the frequency domain;

(c) entering zero values for the amplitudes at the frequencies identified in step (b) in both dimensions to remove any repeating pattern of each of said first and second frequency domain pixel maps said first and second fields of view creating first and second amplitude modified frequency domain pixel maps;

(d) conjugating one of said first and second frequency domain pixel maps of said first and second fields of view from step (a);

(e) multiplying the conjugated one of said first and second frequency domain pixel maps from step (d) with the other one of said first and second frequency domain pixel maps of step (a) that was not conjugated in step (d);

(f) inverse transforming the multiplication result of step (e) using the inverse of the two dimensional Fourier like transformation of step (a) to create the correlation surface of the patterns of said first and second spatial domain pixel maps of said first and second fields of view, said correlation surface is a function of the spatial offset between said patterns;

(g) identifying the peaks of the correlation surface of step (f) along both the x and y axis, the location of those peaks representing the offset, Δ

x and Δ

y, between said first and second spatial domain pixel maps of first and second fields of view;

(h) trigonometrically altering one of said first and second modified frequency domain pixel maps from step (c) using the Δ

x and Δ

y axis offset values of step (g); and

(i) inverse transforming the trigonometrically alerted one of said first and second modified frequency domain pixel maps from step (h) and the non-trigonometericaly altered one of said first and second modified frequency domain pixel maps from step (c) using the inverse of the two dimensional Fourier like transformation of step (a) to obtain first and second modified and aligned spatial domain pixel maps of said first and second fields of view with repeating patterns removed.

2. A method as in claim 1 further comprising the steps of:

(j) subtracting one of said first and second modified and aligned spatial domain pixel maps of step (i) from the other of said first and second modified and aligned spatial domain pixel maps of step (i) to form a difference spatial domain pixel map that only substantially contains any combined noise and defects of said first and second spatial domain pixel maps;

(k) filtering, individually, each of said first and second modified spatial domain pixel domain maps of steps (i) and said difference spatial domain map of step (j) using a threshold amplitude that is above the noise level and below the defect level of each of said first and second modified spatial domain pixel maps and said difference spatial domain pixel map to remove defect shadows introduced by the inverse transformation step (i) where each defect shadow is an integral multiple of each defect with the shadows further being of the opposite color from the real defect and of a lessor amplitude, said filtering producing a first and a second further modified spatial domain pixel map each including random patterns and defects for the corresponding one of said first and second spatial domain pixel maps and a modified difference spatial domain pixel map including only defects for each of said first and second spatial domain pixel maps; and

(l) comparing said first further modified spatial domain pixel map, said second further modified spatial domain pixel map and said modified difference spatial domain pixel map, each from step (k) to determine the defects present in said first spatial domain pixel map in said first field of view.

3. A method as in claim 2 wherein in step (k) the threshold amplitude is selected by an exponential thresholding technique.

4. A method as in claim 2 wherein step (l) includes the steps of:

(m) comparing said first and second further modified spatial domain pixel maps of said first and second fields of view to determine the number of amplitude peaks of each of said first and second further modified spatial domain pixel maps; and

(n) identifying the location of defects as those occurring at the amplitude peaks in said modified difference spatial domain pixel map from step (k) if the number of peaks identified for both said first and second fields of view in step (m) are both large numbers indicating that there are numerous random patterns in said first and second spatial domain pixel maps.

5. A method as in claim 2 wherein step (l) includes the steps of:

(o) comparing said first and second further modified spatial domain pixel maps of said first and second fields of view to determine the number of amplitude peaks of each of said first and second further modified spatial domain pixel maps; and

(p) identifying the location of defects as those occurring at the amplitude peaks in said first further modified spatial domain pixel map of said first field of view from step (k) if the number of peaks identified for both said first and second fields of view in step (o) are both small numbers indicating that there are few or no random patterns in said first and second spatial domain pixel maps.

6. A method as in claim 1 wherein said two dimensional Fourier like transform of step (a) and the inverse transform of step (i) are digital transforms.

7. A method as in claim 1 preceding step (i) further including the step of:

(j) low pass filtering, individually, each of said first and second amplitude modified frequency domain pixel maps from step (c) along both the f_x and f_y axis at frequencies below the lowest frequencies at which zero values were entered in step (c) to remove the harmonic echoes of the random patterns, defects and noise that will occur at values of x and y in the spatial domain below the locations of the first location of the repeating pattern removed in step (c) when step (i) is performed.

8. A method as in claim 1 preceding step (i) further including the step of:

(k) filtering, individually, each of said first and second amplitude modified frequency domain pixel maps from step (c) along both the f_x and f_y axis for compensating the system modular transfer function.

9. An inspection system to detect defects in a device having either random, repeating or mixed patterns thereon comprising:

a processor to control the operation of said inspection system;

first and second optic systems to provide first and second fields of view each under control of said processor and each defines a different optic axis, the first field of view of said first optic system creates a first spatial domain pixel map of a portion of a first device on a first substrate, and said second field of view provides a second spatial domain pixel map against which said first spatial domain pixel map is to be checked from one of a pre-stored pixel map of a desired pattern, a scaled drawing of a desired pattern, a repeating pattern within said first device in said first field of view, a pattern within a second device on said first substrate as in said first field of view, and a second substrate pattern;

a transform system coupled to said first and second optic systems under control of said processor to individually identify and remove repeating patterns from each of said first and second spatial domain pixel maps, said transform system includes;

a two dimensional Fourier like transformer coupled to said first and second optic systems to receive said first and second spatial domain pixel maps to individually transform them into first and second frequency domain pixel maps, respectively;

a harmonic trimmer coupled to said Fourier like transformer to receive said first and second frequency domain pixel maps and to individually determine the frequencies present in said first and second frequency domain pixel maps in both dimensions having amplitude maxima in the frequency domain and individually zeroing out the values at the frequencies at which the maxima values are detected to create first and second modified frequency domain pixel maps, respectively; and

a first two dimensional inverse Fourier like transformer coupled to said harmonic trimmer to receive said first and second modified frequency domain pixel maps to individually transform said first and second modified frequency domain pixel maps using the inverse of a two dimensional Fourier like transform used by said Fourier like transformer to obtain first and second modified spatial domain pixel maps without the repeating patterns of said first and second spatial domain pixel maps;

a conjugator coupled to said Fourier like transformer to receive one of said first and second frequency domain pixel maps and under control of said processor to conjugate the received one of said first and second frequency domain pixel maps;

a multiplier coupled to said conjugator and said Fourier like transformer to receive the other one of said first and second frequency domain pixel maps than that received by said conjugator, under control of said processor, to multiply said conjugated one of first and second frequency domain pixel maps from said conjugator with the other of said first and second frequency domain pixel maps received from said Fourier like transformer to create a multiplied frequency domain pixel map;

a second inverse Fourier like transformer couple to said multiplier to receive said multiplied frequency domain pixel map to inverse transform said multiplied frequency domain pixel map using the inverse of the Fourier like transform used by said Fourier like transformer to create a correlation surface of the patterns of said first and second spatial domain pixel maps, said correlation surface being a function of the spatial offset between said first and second spatial domain pixel maps;

a peak detector coupled to said second inverse Fourier like transformer to receive said correlation surface to identify the peaks of said correlation surface along both the x and y axis with the location of those peaks representing the offset, Δ

x and Δ

y, between said first and second spatial domain pixel maps; and

a shifting filter coupled to said peak detector to receive said offset, Δ

X and Δ

y, and said harmonic trimmer to receive one of said first and second modified frequency domain pixel map to trigonometrically alter the two dimensional sinusoids of the received one of said first and second modified frequency domain pixel maps using the Δ

x and Δ

y axis offset values identified by said peak detector to align the received one of said first and second modified frequency domain pixel maps with the other of said first and second modified frequency domain pixel maps.

10. In an inspection system for detecting defects in a device having either random, repeating or mixed patterns thereon, said inspection system includes first and second fields of view each positioned along a different optic axis, the first field of view creates a first spatial domain pixel map of a portion of a first device on a first substrate;

and said second field of view provides a second spatial domain pixel map against which said first spatial domain pixel map is to be checked;

said second spatial domain pixel map is developed from one of a pre-stored pixel map of a desired pattern, a scaled drawing of a desired pattern, a repeating pattern within said first device in said first field of view, a pattern within a second device on said first substrate as in said first field of view, and a pattern on a second substrate;

a method for detecting defects in said first device in said first field of view comprising the steps of;

(a) transforming said first and second spatial domain pixel maps from said first and second fields of view using a two dimensional Fourier like transform to convert said first and second spatial domain pixel maps from the spatial domain to first and second frequency domain pixel maps, respectively;

(b) identifying those frequencies in both dimensions having amplitude maxima in the frequency domain of said first and second frequency domain pixel maps with any repeating pattern of each of said first and second frequency domain pixel maps from said first and second fields of view occurring at those frequencies in the frequency domain;

(c) entering zero values for the amplitudes at the frequencies identified in step (b) in both dimensions to remove any repeating pattern of each of said first and second frequency domain pixel maps said first and second fields of view creating first and second amplitude modified frequency domain pixel maps;

(d) inverse transforming said first and second amplitude modified frequency domain pixel maps from step (c) using the inverse of the two dimensional Fourier like transformation of step (a) to obtain first and second modified spatial domain pixel maps that are substantially equivalent to said first and second spatial domain pixel maps of said first and second fields of view with any repeating patterns removed;

(e) subtracting one of said first and second modified spatial domain pixel maps of step (d) from the other of said first and second modified spatial domain pixel maps of step (d) to form a difference spatial domain pixel map that only substantially contains any combined noise and defects of said first and second spatial domain pixel maps;

(f) filtering, individually, each of said first and second modified spatial domain pixel maps of step (d) and said difference spatial domain map of step (e) using a threshold amplitude that is above the noise level and below the defect level of each of said first and second modified spatial domain pixel map and said difference spatial domain pixel map to remove defect shadows introduced by the inverse transformation step (d) where each defect shadow is an integral multiple of each defect with the shadows further being of the opposite color from a real defect and of a lessor amplitude, said filtering producing a first and a second further modified spatial domain pixel map each including any random pattern and defects for the corresponding one of said first and second spatial domain pixel maps, and a modified difference spatial domain pixel map including only defects for each of said first and second spatial domain pixel maps; and

(g) comparing said first further modified spatial domain pixel map, said second further modified spatial domain pixel map and said modified difference spatial domain pixel map, each from step (f), to determine the defects present in said first spatial pixel map in said first field of view.

11. A method as in claim 10 wherein in step (f) the threshold amplitude is selected by an exponential thresholding technique.

12. A method as in claim 10 wherein step (g) includes the steps of:

(h) comparing said first and second further modified spatial domain pixel maps of said first and second fields of view to determine the number of amplitude peaks of each of said first and second spatial domain pixel maps; and

(i) identifying the location of defects as those occurring at the amplitude peaks in said modified difference spatial domain pixel map from step (f) if the number of peaks identified for both said first and second fields of view in step (h) are both large numbers indicating that there are numerous random patterns in said first and second spatial domain pixel maps.

13. A method as in claim 10 wherein step (g) includes steps of:

(j) comparing said first and second further modified spatial domain pixel maps of said first and second fields of view to determine the number of amplitude peaks of each of said first and second spatial domain pixel maps; and

(k) identifying the location of defects as those occurring at the amplitude peaks in said first further modified spatial domain pixel map of said first field of view from step (f) if the number of peaks identified for both said first and second fields of view in step (j) are both small numbers indicating that there are few or no random patterns in the original first and second spatial domain pixel maps.

14. A method as in claim 10 wherein said two dimensional Fourier like transform of step (a) and the inverse transform of step (d) are digital transforms.

15. A method as in claim 10 preceding step (d) further including the step of:

(h) low pass filtering, individually, each of said first and second fields of view from step (c) along both the f_x and f_y axis at frequencies below the lowest frequencies at which zero values were entered in step (c) to remove the harmonic echoes of the random patterns, defects and noise that will occur at values of x and y in the spatial domain below the locations of the first location of the repeating pattern removed in step (c) when step (d) is performed.

16. A method as in claim 10 preceding step (d) further including the step of:

(i) filtering, individually, each of said first and second fields of view from step (c) along both the f_x and f_y axis for compensating the system modular transfer function.

17. In an inspection system for detecting defects in a device having either random, repeating or mixed patterns thereon, said inspection system includes first and second fields of view each positioned along a different optic axis, the first field of view creates a first spatial domain pixel map of a portion of a first device on a first substrate;

and said second field of view provides a second spatial domain pixel map against which said first spatial domain pixel map is to be checked;

said second spatial domain pixel map is developed from one of a pre-stored pixel map of a desired pattern, a scaled drawing of a desired pattern, a repeating pattern within said first device in said first field of view, a pattern within a second device on said first substrate as in said first field of view, and a pattern on a second substrate;

a method for detecting defects in said first device in said first field of view comprising the steps of;

(a) transforming said first and second spatial domain pixel maps from said first and second fields of view using a two dimensional Fourier like transform to convert said first and second spatial domain pixel maps from the spatial domain to first and second frequency domain pixel maps, respectively;

(b) identifying those frequencies in both dimensions having amplitude maxima in the frequency domain of said first and second frequency domain pixel maps with any repeating pattern of each of said first and second frequency domain pixel maps from said first and second fields of view occurring at those frequencies in the frequency domain;

(c) entering zero values for the amplitudes at the frequencies identified in step (b) in both dimensions to remove any repeating pattern of each of said first and second frequency domain pixel maps said first and second fields of view creating first and second amplitude modified frequency domain pixel maps;

(d) subtracting one of said first and second amplitude modified frequency domain pixel maps of step (c) from the other of said first and second amplitude modified frequency domain pixel maps of step (c) to form a difference frequency domain pixel map that contains only the combined noise and defects of said first and second frequency domain pixel maps;

(e) individually inverse transforming said first and second amplitude modified frequency domain pixel maps from step (c) using the inverse of the two dimensional Fourier like transformation of step (a) to obtain first and second modified spatial domain pixel maps of said first and second fields of view with any repeating patterns of said first and second spatial domain pixel maps removed;

(f) inverse transforming said difference frequency domain pixel map from step (d) using the inverse of the two dimensional Fourier like transformation of step (a) to obtain a difference spatial domain pixel map that contains only the combined noise and defects of said first and second spatial domain pixel maps;

(g) filtering, individually, each of said first and second modified spatial domain pixel maps of step (e) and said difference spatial domain map of step (f) using a threshold amplitude that is above the noise level and below the defect level of each of said first and second modified spatial domain pixel maps and said difference spatial domain map to remove defect shadows introduced by the inverse transformation steps (e) and (f) where each defect shadow is an integral multiple of each defect with the shadows further being of the opposite color from the real defect and of a lessor amplitude, said filtering producing a first and a second further modified spatial domain pixel map each including any random pattern and defects for the corresponding one of said first and second spatial domain pixel maps, and a modified difference spatial domain pixel map including only defects for each of said first and second spatial domain pixel maps; and

(h) comparing said first further modified spatial domain pixel map, said second further modified spatial domain pixel map and said modified difference spatial domain pixel map, each from step (g), to determine the defects present in said first spatial domain pixel map in said first field of view.

18. A method as in claim 17 wherein in step (g) the threshold amplitude is selected by an exponential thresholding technique.

19. A method as in claim 17 wherein step (h) includes the steps of:

(i) comparing said first and second further modified spatial domain pixel maps of said first and second fields of view to determine the number of amplitude peaks of each of said first and second further modified spatial domain pixel maps; and

(j) identifying the location of defects as those occurring at the amplitude peaks in said modified difference spatial domain pixel map from step (g) if the number of peaks identified for both said first and second field of view in step (i) are both large numbers indicating that there are numerous random patterns in the original first and second spatial domain pixel maps.

20. A method as in claim 17 wherein step (h) includes the steps of:

(k) comparing said first and second further modified spatial domain pixel maps of said first and second fields of view to determine the number of amplitude peaks of each of said first and second further modified spatial domain pixel maps; and

(l) identifying the location of defects as those occurring at the amplitude peaks in said first further modified spatial domain pixel map of said first field of view from step (g) if the number of peaks identified for both said first and second fields of view in step (g) are both small numbers indicating that there are few or no random patterns in said first and second spatial domain pixel maps.

21. A method as in claim 17 wherein said two dimensional Fourier like transform of step (a) and the inverse transform of step (f) are digital transforms.

22. A method as in claim 17 preceding step (f) further including the step of:

(i) low pass filtering, individually, each of said first and second fields of view from step (c) along both the f_x and f_y axis at frequencies below the lowest frequencies at which zero values were entered in step (c) to remove the harmonic echoes of the random patterns, defects and noise that will occur at values of x and y in the spatial domain below the locations of the first location of the repeating pattern removed in step (c) when step (f) is performed.

23. A method as in claim 17 preceding step (f) further including the step of:

(j) filtering, individually, each of said first and second fields of view from step (c) along both the f_x and f_x axis for compensating the system modular transfer function.

24. An inspection system to detect defects in a device having either random, repeating or mixed patterns thereon comprising:

a processor to control the operation of said inspection system;

first and second optic systems to provide first and second fields of view each under control of said processor and each defines a different optic axis, the first field of view of said first optic system creates a first spatial domain pixel map of a portion of a first device on a first substrate, and said second field of view provides a second spatial domain pixel map against which said first spatial domain pixel map is to be checked from one of a pre-stored pixel map of a desired pattern, a scaled drawing of a desired pattern, a repeating pattern within said first device in said first field of view, a pattern within a second device on said first substrate as in said first field of view, and a second substrate pattern;

a transform system coupled to said first and second optic systems under control of said processor to identify and remove repeating patterns from each of said first and second spatial domain pixel maps, said transform system includes;

a two dimensional Fourier like transformer coupled to said first and second optic systems to receive said first and second spatial domain pixel maps to individually transform them into first and second frequency domain pixel maps, respectively;

a harmonic trimmer coupled to said Fourier like transformer to receive said first and second frequency domain pixel maps and to individually determine the frequencies present in said first and second frequency domain pixel maps in both dimensions having amplitude maxima in the frequency domain and individually zeroing out the values at the frequencies at which the maxima values are detected to create first and second modified frequency domain pixel maps, respectively; and

a first two dimensional inverse Fourier like transformer coupled to said harmonic trimmer to receive said first and second modified frequency domain pixel maps to individually transform said first and second modified frequency domain pixel maps using the inverse of a two dimensional Fourier like transform used by said Fourier like transformer to obtain first and second modified spatial domain pixel maps without the repeating patterns of said first and second spatial domain pixel maps;

a subtractor coupled to said first inverse Fourier like transformer to receive said first and second modified spatial domain pixel maps under control of said processor to subtract one of said first and second modified spatial domain pixel maps from the other of said first and second modified spatial domain pixel maps to form a difference spatial domain pixel map that only substantially contains the combined noise and defects of both of said first and second modified spatial domain pixel maps;

first, second and third threshold filters, said first and second threshold filters individually connected to said first inverse Fourier like transformer to received said first and second modified spatial domain pixel maps, respectively, and said third threshold filter coupled to said subtractor to receive said difference spatial domain pixel map, said first second and third threshold filters to each individually filter said first modified spatial domain pixel map, said second modified spatial domain pixel map and said difference spatial domain pixel map using a threshold amplitude that is selected to be above the noise level and below the defect level, to remove defect shadows introduced by said first and second inverse Fourier like transformers where each defect shadow is an integral multiple of each defect with the shadows further being of the opposite color from the real defect and of a lessor amplitude, to individually create first and second further modified spatial domain pixel maps and a modified difference spatial domain pixel map, respectively, said first and second further modified spatial domain pixel maps each includes the random and defects for each of said first and second spatial domain pixel maps, respectively, and said modified difference spatial domain pixel map contains only defects for each of said first and second spatial dominant pixel maps; and

defect analyzer coupled to said first, second and third threshold filters to receive said first and second further modified spatial domain pixel maps and said modified difference spatial domain pixel map from which to compare said first further modified spatial domain pixel map, said second further modified spatial domain pixel map and said modified difference spatial domain pixel map to determine the defects present in said first spatial domain pixel map.

25. An inspection system as in claim 24 wherein in each of said first, second and third threshold filters the threshold amplitude is selected by an exponential thresholding technique.

26. An inspection system as in claim 24 wherein said defect analyzer includes:

a comparator coupled to said first, second and third threshold filters, under control of said processor, to receive said first and second further modified spatial domain pixel maps and said modified difference spatial domain pixel map to compare said three received pixel maps to determine the number of amplitude peaks of each of said received pixel maps; and

a peak detector coupled to said first, second and third threshold filters and said comparator, under control of said processor, to receive said first and second further modified spatial domain pixel maps and said modified difference spatial domain pixel map, and comparison results;

said peak detector to detect and identify the location of defects in each of the received pixel maps, said defects being identified as those occurring at the amplitude peaks in said modified difference spatial domain pixel map if the number of peaks identified for both of said first and second further modified spatial domain pixel maps by said comparator are both large numbers indicating that there are numerous random patterns in said first and second spatial domain pixel maps.

27. An inspection system as in claim 24 wherein defect analyzer includes:

a comparator coupled to said first, second and third threshold filters, under control of said processor, to receive said first and second further modified spatial domain pixel maps and said modified difference spatial domain pixel map to compare said three received pixel maps to determine the number of amplitude peaks of each of said received pixel maps; and

a peak detector coupled to said first, second and third threshold filters and said comparator, under control of said processor, to receive said first and second further modified spatial domain pixel maps and said modified difference spatial domain pixel map, and comparison results;

said peak detector to detect and identify the location of defects in each of the received pixel maps, said defects being identified as those occurring at the amplitude peaks in said first and second further modified spatial domain pixel maps if the number of peaks identified for both of said first and second further modified spatial domain pixel maps by said comparator are both small numbers indicating that there are few or no random patterns in said first and second spatial domain pixel maps.

28. An inspection system to detect defects in a device having either random, repeating or mixed patterns thereon comprising:

a processor to control the operation of said inspection system;

first and second optic systems to provide first and second fields of view each under control of said processor and each defines a different optic axis, the first field of view of said first optic system creates a first spatial domain pixel map of a portion of a first device on a first substrate, and said second field of view provides a second spatial domain pixel map against which said first spatial domain pixel map is to be checked from one of a pre-stored pixel map of a desired pattern, a scaled drawing of a desired pattern, a repeating pattern within said first device in said first field of view, a pattern within a second device on said first substrate as in said first field of view, and a second substrate pattern;

a transform system coupled to said first and second optic systems under control of said processor to identify and remove repeating patterns from each of said first and second spatial domain pixel maps, said transform system includes;

a two dimensional Fourier like transformer coupled to said first and second optic systems to receive said first and second spatial domain pixel maps to individually transform them into first and second frequency domain pixel maps, respectively;

a harmonic trimmer coupled to said Fourier like transformer to receive said first and second frequency domain pixel maps and to individually determine the frequencies present in said first and second frequency domain pixel maps in both dimensions having amplitude maxima in the frequency domain and individually zeroing out the values at the frequencies at which the maxima values are detected to create first and second modified frequency domain pixel maps, respectively; and

a first two dimensional inverse Fourier like transformer coupled to said harmonic trimmer to receive said first and second modified frequency domain pixel maps to individually transform said first and second modified frequency domain pixel maps using the inverse of a two dimensional Fourier like transform used by said Fourier like transformer to obtain first and second modified spatial domain pixel maps without the repeating patterns of said first and second spatial domain pixel maps;

a subtractor coupled to said harmonic trimmer to receive said first and second modified frequency domain pixel maps under control of said processor to subtract one of said first and second modified frequency domain pixel maps from the other of said first and second modified frequency domain pixel maps to form a difference frequency domain pixel map that only substantially contains the combined noise and defects of both of said first and second modified frequency domain pixel maps; and

a second two dimensional inverse Fourier like transformer, using the inverse of a two dimensional Fourier like transform used by said Fourier like transformer, coupled to said subtractor to receive said difference frequency domain pixel map under control of said processor to transform said difference frequency domain pixel map to the spatial domain as a difference spatial domain pixel map;

first, second and third threshold filters, said first and second threshold filters individually connected to said first inverse Fourier like transformer to received said first and second modified spatial domain pixel maps, respectively, and said third threshold filter coupled to said second inverse Fourier like transformer to receive said difference spatial domain pixel map, said first second and third threshold filters to each individually filter said first modified spatial domain pixel map, said second modified spatial domain pixel map and said difference spatial domain pixel map using a threshold amplitude that is selected to be above the noise level and below the defect level, to remove defect shadows introduced by said first and second inverse Fourier like transformers where each defect shadow is an integral multiple of each defect with the shadows further being of the opposite color from the real defect and of a lessor amplitude, to individually create first and second further modified spatial domian pixel maps and a modified difference spatial domain pixel map, respectively, said first and second further modified spatial domain pixel maps each includes the random and defects for each of said first and second spatial domain pixel maps, respectively, and said modified difference spatial domain pixel map contains only defects for each of said first and second spatial domain pixel maps; and

defect analyzer coupled to said first, second and third threshold filters to receive said first and second further modified spatial domain pixel maps and modified difference spatial domain pixel map to compare said first further modified spatial domain pixel map, said second further modified spatial domain pixel map and said modified difference spatial domain pixel map to determine the defects present in said first spatial domain pixel map.

29. An inspector system as in claim 28 wherein said first, second and third threshold filters the threshold amplitude is selected by an exponential thresholding technique.

30. An inspection system as in claim 28 wherein defect analyzer includes:

comparator coupled to said first, second and third threshold filters, under control of said processor, to receive said first and second further modified spatial domain pixel maps and said modified difference spatial domain pixel map to compare said three received pixel maps to determine the number of amplitude peaks of each of said received pixel maps; and

a peak detector coupled to said first, second and third threshold filters and said comparator, under control of said processor, to receive said first and second further modified spatial domain pixelmaps and said modified difference spatial domain pixel map, and comparison results;

said peak detector to detect and identify the location of defects in each of the received pixel maps, said defects being identified as those occurring at the amplitude peaks in said first and second further modified spatial domain pixel maps if the number of peaks identified for both of said first and second further modified spatial domain pixel maps by said comparator are both small numbers indicating that there are few or no random patterns in said first and second spatial domain pixel maps.