JP6252906B2 - Opening pattern design method, image restoration method, and image restoration apparatus - Google Patents

Description

  The present invention relates to a technique for designing an opening pattern used for image restoration.

  In an inspection device, an alignment device, and the like, an image pickup unit that picks up an object generally has an autofocus function. Thereby, it becomes possible to acquire an image by aligning the in-focus position with the surface of the object. However, if the autofocus operation that automatically adjusts the focus position to the surface of the object is not performed properly for some reason, imaging is performed with the focus position shifted from the surface of the object. A blurred image that is a focused image is acquired.

  When the blurred image is acquired, for example, the inspection apparatus cannot perform inspection accurately. When a blurred image of the alignment mark is acquired by the drawing apparatus, there arises a problem that the alignment of the object fails or the drawing accuracy is lowered.

  Therefore, development of a technique for restoring a true image from a blurred image is underway. In the image restoration, a focused image is estimated by performing a deconvolution of a blurred image and a point spread function (PSF) corresponding to a defocus amount at the time of imaging. At this time, noise called ringing is likely to occur in the restored image. Therefore, it has been proposed to provide an aperture stop having a special aperture pattern called a coded aperture or a coded stop. Ringing can be reduced by changing the PSF with the coded aperture and increasing the accuracy of deconvolution (see, for example, Patent Documents 1 and 2). Non-Patent Document 1 discloses an opening pattern for improving the restoration accuracy of an image under the assumption that the power spectrum, which is the frequency characteristic of a natural image, follows 1 / f.

JP 2010-81460 A JP 2012-22308 A

Changyin Zhou, 1 other, “What are Good Apertures for Defocus Deblurring?”, [Online], 2009, Columbia University, [April 17, 2013 search], Internet <URL: http: //www1.cs .columbia.edu / ~ changyin / files / ZhouICCP2009.pdf>

  By the way, conventionally, research on image restoration has been performed on natural images. In this case, the shape of the PSF substantially follows the shape of the coding aperture. Therefore, as shown in Non-Patent Document 1, when obtaining the pattern of the coded aperture, the calculation has been performed on the assumption that the aperture pattern and the PSF are the same. Specifically, the PSF is treated as the same binary pattern as the coded aperture.

  However, the PSF is actually a multi-value image pattern, and due to the difference between the binary PSF and the multi-value PSF, sufficient image restoration performance is not exhibited even if the obtained coded aperture is used. There is. In particular, unlike natural image capturing, when image capturing is performed with light having a narrow wavelength band, the difference between the coded aperture and the PSF increases, and the image restoration performance is significantly degraded.

  The present invention has been made in view of the above problems, and an object of the present invention is to realize good image restoration even when imaging is performed with light in a narrow wavelength band.

  The invention described in claim 1 is a method for designing an opening pattern used in an imaging unit of an image restoration device, wherein a) a step of preparing a plurality of candidate patterns which are opening patterns, and b) a predetermined method. A step of acquiring a point spread function obtained by the imaging unit through each of the plurality of candidate patterns and having an Airy pattern as a multi-gradation image when a point light source is disposed at a predetermined position. And c) the accuracy when the blurred image obtained when the imaging target is arranged at the predetermined position is restored using each of the plurality of point spread functions obtained in the step b). A step of obtaining an evaluation value; and d) a step of selecting one of the plurality of candidate patterns based on the plurality of evaluation values obtained in the step c).

  The invention according to claim 2 is the opening pattern design method according to claim 1, wherein in the step b), the point spread function is a light beam for light emitted from a virtual point light source. Obtained by tracking.

  A third aspect of the present invention is the opening pattern design method according to the first or second aspect, wherein the evaluation value obtained in step c) indicates a plurality of imaging targets belonging to one category. It is obtained using a representative spectrum obtained from the spectrum of the image.

The invention according to claim 4 is the opening pattern design method according to claim 3, wherein the frequency is ξ, the representative spectrum is S _ξ , and the PSF corresponding to the candidate pattern is subjected to Fourier transform, K _ξ , With the noise level as σ, the evaluation value R of the candidate pattern is obtained by Equation 1.

  The invention according to claim 5 is an image restoration method, wherein the imaging is provided with an aperture stop having an aperture pattern determined by the aperture pattern design method according to any one of claims 1 to 4. Capturing an image of an imaging target irradiated with light having the same wavelength or the same wavelength band as the light emitted from the point light source, and a multi-tone point spread function corresponding to the opening pattern And a step of acquiring a restored image from the image.

  An invention according to claim 6 is an image restoration apparatus, and an imaging unit provided with an aperture stop having an aperture pattern determined by the aperture pattern design method according to any one of claims 1 to 4; Corresponding to the aperture pattern from an illumination unit that irradiates the imaging target with light having the same wavelength or the same wavelength band as the light emitted from the point light source, and an image acquired by imaging the imaging target by the imaging unit An image restoration unit that obtains a restored image using a multi-gradation point spread function.

  A seventh aspect of the present invention is the image restoration device according to the sixth aspect, wherein the illumination unit uses a laser, a light emitting diode, or a combination of a white light source and a narrow bandpass filter as a light source. Including.

  The invention according to claim 8 is the image restoration device according to claim 6 or 7, wherein the reference of the imaging target is obtained by pattern matching between the restored image of the imaging target that is an alignment mark and a template image. A positional deviation amount obtaining unit for obtaining a positional deviation amount from the position is further provided.

  According to the present invention, good image restoration can be realized even when imaging is performed with light of a narrow wavelength band.

It is a figure which shows the structure of the alignment apparatus. It is a block diagram which shows the function structure which a computer implement | achieves. It is a figure which shows the flow of the process which concerns on the design of an opening pattern. It is a figure which shows the flow of an opening pattern determination process. It is a figure which shows the example of an optimization pattern. It is a figure which shows the flow of the alignment process of a board | substrate. It is a figure which shows an alignment mark. It is a figure which shows the example of an opening pattern. It is a figure which shows the example of PSF which is a multi-value image. It is a figure which shows the power spectrum of an opening pattern and PSF. It is a figure which shows the example of a blurred image. It is a figure which shows the decompression | restoration image which has ringing. It is a figure which shows PSF by white light. It is a figure which shows PSF containing an Airy pattern. It is a figure which shows the error of the detection position of an alignment mark. It is a figure which shows the example of a focused image. It is a figure which shows the blur image acquired via the circular aperture stop. It is a figure which shows the image decompress | restored using binary PSF. It is a figure which shows the blur image acquired via the aperture stop by a prior art. It is a figure which shows the image decompress | restored using binary PSF. It is a figure which shows the blur image acquired through the aperture stop concerning this Embodiment. It is a figure which shows the image decompress | restored using multi-gradation PSF. It is a figure which shows the example of an opening pattern. It is a figure which shows PSF actually acquired. It is a figure which shows PSF calculated | required by calculation. FIG. A thru | or FIG. It is a figure which shows the power spectrum of C.

  FIG. 1 is a diagram showing a configuration of an alignment apparatus 1 according to an embodiment of the present invention. The alignment apparatus 1 is an apparatus that arranges a substrate 9 such as a semiconductor substrate or a glass substrate at a reference position. The alignment device 1 is provided in an inspection device, a drawing device, and the like, and the substrate 9 is arranged at a reference position, thereby realizing a pattern inspection in the inspection device, a pattern drawing in the drawing device, and the like with high accuracy. Is done.

  The alignment apparatus 1 receives an image pickup unit 2 that picks up an alignment mark formed on the substrate 9, a moving mechanism 3 that moves the substrate 9 relative to the image pickup unit 2, and image data from the image pickup unit 2. A computer 4 is provided. The imaging unit 2 electrically outputs an illumination unit 21 that emits illumination light, an optical system 22 that guides illumination light to the substrate 9 and receives light from the substrate 9, and an image of the substrate 9 formed by the optical system 22. It has an image sensor 23 (line sensor or area sensor) that converts it into a signal. The illumination unit 21 includes a laser as a light source. The illumination unit 21 may include a light emitting diode or a combination of a white light source and a narrow bandpass filter as a light source. The wavelength range of illumination light is narrow. A combination of a light emitting diode and a narrow bandpass filter may be used for the light source.

  The optical system 22 has an aperture stop 221 in which an aperture pattern described later is formed. The aperture stop 221 is also referred to as a “coded aperture”, and the aperture pattern is also referred to as a “coded aperture”. The moving mechanism 3 includes a stage 31 that holds the substrate 9, a rotating mechanism 32 that rotates the stage 31 around an axis parallel to the Z direction in FIG. 1, and a Y direction moving mechanism 33 that moves the stage 31 in the Y direction. And an X-direction moving mechanism 34 for moving the stage 31 in the X direction. The X direction, the Y direction, and the Z direction are perpendicular to each other.

  FIG. 2 is a block diagram showing a functional configuration realized by the computer 4. In FIG. 2, the imaging unit 2 and the moving mechanism 3 are also indicated by broken-line blocks. The computer 4 includes a calculation unit 41 that performs various calculations and a control unit 42 that performs overall control of the alignment apparatus 1. The calculation unit 41 includes an image restoration unit 411 and a positional deviation amount acquisition unit 412 described later. The function of the calculation unit 41 may be constructed by a dedicated electric circuit, or a dedicated electric circuit may be partially used.

  As described above, in the imaging unit 2 of the alignment apparatus 1, the aperture stop 221 having an aperture pattern having a shape corresponding to the imaging target on the substrate 9 is designed and manufactured in advance. Further, PSF (point spread function) which is information used in the alignment process of the substrate 9 described later is also acquired in advance corresponding to the opening pattern. Hereinafter, the process related to the opening pattern design, that is, the opening pattern design and manufacturing process, and the point spread function acquisition process performed as a preliminary preparation for the alignment process of the substrate 9 will be described with reference to FIG. To do.

  In the process related to the opening pattern design, first, a plurality of images showing a plurality of types of alignment marks having different shapes are prepared (step S11). Here, a plurality of substrates on which a plurality of types of alignment marks are formed (hereinafter referred to as “reference substrates” in order to be distinguished from a substrate 9 in an alignment process described later) are prepared. In the alignment apparatus 1, the alignment marks on the plurality of reference substrates are imaged, and a plurality of images (signals thereof) are acquired.

  At this time, the focus function in the imaging unit 2 is adjusted so that the alignment mark is focused. In addition, as the aperture of the aperture stop 221 in the imaging unit 2, an aperture having a general substantially circular aperture region is used. Each alignment mark includes a plurality of linear portions extending in directions orthogonal to each other and has anisotropy. In a plurality of images, these linear portions are oriented in substantially the same direction. Note that a plurality of images indicating a plurality of types of alignment marks may be acquired by other imaging units. Further, the surface of the reference substrate may be focused by providing a mechanism for moving the stage 31 in the Z direction.

  When a plurality of images are prepared, the calculation unit 41 acquires a power spectrum for each direction in the image by performing a Fourier transform on each of the plurality of images. Thereby, a plurality of power spectra in a plurality of images are acquired for each direction (step S12). Subsequently, for each direction, a representative value of the values of the plurality of power spectra at each frequency is obtained. Here, the representative value is a value indicating the vicinity of the center of the range in which the values of a plurality of power spectra at each frequency are distributed, such as an average value or a median value. And the spectrum expressed by the representative value in each frequency is acquired as a representative spectrum in each direction (step S13).

  The representative spectrum can be regarded as representing the power spectrum of a plurality of images in each direction. The representative spectrum in each direction is acquired by acquiring a plurality of power spectrum images from the plurality of images prepared in step S11 and obtaining representative values of pixel values corresponding to each other in the plurality of power spectrum images. Good.

  Further, in the alignment apparatus 1, a noise level (here, standard deviation of noise) in the image pickup device 23 of the image pickup unit 2 is measured (step S14). For example, the noise level is measured by acquiring the output from the image sensor 23 in a state where the objective lens of the imaging unit 2 is provided with a lid that shields light. The noise level may be obtained by calculation from an image acquired by the imaging unit 2 (for example, an image acquired by the process of step S11).

  Subsequently, an aperture pattern used in the imaging unit 2 is determined using the noise level and the representative spectrum in the imaging unit 2 (step S15). In the opening pattern determination process in the present embodiment, a method using a genetic algorithm described in Table 1 of Non-Patent Document 1 is used.

  FIG. 4 is a diagram showing the flow of the opening pattern determination process, and shows the process performed in step S15 of FIG. Here, the shape of the opening pattern is regarded as a matrix of N rows and N columns in which “1” or “0” is assigned to each element. “1” corresponds to the opening region, and “0” corresponds to the light shielding region. In this embodiment, N is 13, but the value of N can be variously changed. In the following description, the matrix is represented by a binary sequence (hereinafter referred to as “sequence”) having a length (N × N). The sequence and the opening pattern are substantially synonymous.

  In the opening pattern determination process, first, “1” or “0” is irregularly assigned to (N × N) element positions of a sequence, thereby making R sequences different from each other (hereinafter referred to as “initial sequence”). .) Is generated (step S21). Next, the PSF when provisionally obtained using the opening pattern indicated by the initial sequence is obtained as a multi-gradation image (hereinafter referred to as “multi-valued image”) by calculation using a ray tracing method or the like (step S22). ). That is, when a point light source is virtually arranged at a predetermined position, the PSF acquired by the imaging unit 2 via the opening pattern is acquired as a multi-value image.

  For example, the optical design software ZEMAX (ZEMAX, Washington, USA) can be used to input the light wavelength, lens data (lens material and diameter, distance between lenses, etc.), aperture shape of the aperture, etc. Desired. Of course, other software such as codeV (synopsys, California, USA) may be used, and the PSF may be obtained by other simulation methods.

Subsequently, when an image is captured using the opening pattern indicated by each initial sequence, the restoration accuracy of the image restored from the image (for example, the difference between the true image and the restored image) is shown. An evaluation value is obtained for each direction (step S23). Specifically, the frequency is ξ, the representative spectrum in each direction is S _ξ , the PSF corresponding to each initial sequence is Fourier transformed with respect to each direction, K _ξ , and the noise level of the imaging unit 2 is σ. The evaluation value R in each direction of the initial sequence is obtained using Equation 2 according to Equation 12 of Non-Patent Document 1.

In Non-Patent Document 1, since an object is a natural image illuminated with white light, it is considered that the aperture pattern and the PSF match. Therefore, calculation of evaluation values with those of the opening pattern and the Fourier transform as K _xi] is performed. However, in the present embodiment, the PSF is assumed to be a multi-valued image, the calculation of Equation 2 is performed, and the evaluation value is obtained.

  Thereafter, a sum of evaluation values in a plurality of directions (more precisely, all directions to be calculated) for each initial sequence is further obtained, and 1st to Mth (where M is 2) among the R initial sequences. Each of the M initial sequences having a small sum of evaluation values is selected as a candidate sequence (step S24).

  Subsequently, of the M candidate sequences, two candidate sequences are randomly selected and duplicated to prepare two duplicate sequences (step S25). Element positions in the sequence are selected with a predetermined probability c1 (for example, 0.2), and the values of the element positions in the two duplicate sequences are exchanged to obtain two new sequences. That is, “crossover” is performed to exchange values at the same position in two number sequences (step S26). In each of the two sequences, the element position of the sequence is selected with a predetermined probability c2 (for example, 0.05), and the value “1” of the element position is set to “0” or “0” is set. Inverted to “1”. That is, “mutation” is performed to change the value of the selected element position. The generated two new candidate sequences are added to the M candidate sequences (step S27). Steps S25 to S27 are repeated until the number of candidate sequences becomes R (step S28).

  Subsequently, the processes in steps S22 to S28 are repeated with the R candidate sequences acquired in the immediately preceding step S27 as R initial sequences (step S29). In other words, the generation is updated and the process is repeated. The process in step S29 is repeated (G-1) times in advance, with the R candidate sequences acquired in the immediately preceding step S27 as R initial sequences. Thereby, (R × G) candidate sequences, that is, (R × G) candidate patterns of the opening pattern are prepared.

  Then, as in step S23, the evaluation value R for each direction of the (R × G) candidate patterns is obtained from Equation 2. However, the evaluation value may be used for the candidate pattern for which the evaluation value has already been obtained. For each candidate pattern, the sum of evaluation values in all directions is obtained as a final evaluation value, and one candidate pattern with the lowest final evaluation value is an aperture pattern used in the imaging unit 2. It determines as an optimization pattern to show (step S30). The final evaluation value may be another value based on the evaluation values in all directions, for example, an average value, a median value, a maximum value, or a minimum value. The same applies to the selection of candidate sequences in step S24.

  FIG. 5 is a diagram illustrating an opening pattern 8 which is an example of an optimization pattern. The dark square indicates the element 81 in the opening area, and the white square indicates the element 82 in the light shielding area. The aperture stop 221 is produced according to the aperture pattern (step S16). The aperture stop 221 is attached to the optical system 22 of the imaging unit 2 in place of the aperture stop having a circular aperture region (step S17).

  In the alignment apparatus 1, a multivalued image of a point light source is acquired as a PSF using the imaging unit 2 (step S18). Here, the PSF is acquired for each of the plurality of defocus amounts, with the distance on the optical axis between the point light source that is the imaging target and the ideal in-focus position as the defocus amount. The PSF is stored in the calculation unit 41. The plurality of PSFs may be obtained by calculation using a ray tracing method or the like. With the above process, the process related to the opening pattern design, which is a preliminary preparation for the alignment process of the substrate 9, is completed.

  The acquisition of the representative power spectrum and the determination of the aperture pattern in the above processing may be executed by another computer different from the computer 4.

  FIG. 6 is a diagram showing the flow of the alignment process for the substrate 9 in the alignment apparatus 1. In the alignment apparatus 1, first, the substrate 9 placed on the stage 31 is imaged by the imaging unit 2, and an image (hereinafter referred to as “target image”) is acquired (step S41). At this time, the illumination unit 21 emits light in the same wavelength band as the wavelength band used when acquiring the PSF. Of course, this wavelength band is the same as the wavelength band set when the aperture pattern was designed. The target image is output to the calculation unit 41 of the computer 4. For example, the alignment mark 91 shown in FIG. 7 is formed on the substrate 9, and the imaging region on the substrate 9 by the imaging unit 2 includes the alignment mark 91 that is an imaging target. That is, in the process of step S41, a target image indicating an imaging target of the same category as the plurality of images prepared in the process of step S11 of FIG. 3 is acquired.

  The image restoration unit 411 determines whether or not the target image is a blurred image based on, for example, the contrast of the target image, the spread of the histogram of pixel values, and the like. When the target image is a blurred image, a restored image is acquired from the target image using the PSF prepared in advance in step S18 (step S42). In obtaining the restored image, for example, the well-known Wiener deconvolution is used. Here, a plurality of restored image candidates are acquired by Wiener deconvolution while changing each of the defocus amount and the noise level in a plurality of ways, and the one with the highest contrast among the plurality of restored image candidates is determined as the restored image. Is done. That is, the defocus amount and the noise level are acquired simultaneously with the acquisition of the restored image.

  In acquiring the restored image, the noise level acquired in step S14 may be used. Further, the alignment apparatus 1 may be provided with a laser length measuring device, and a restored image may be acquired using a defocus amount obtained from a measurement result of the laser length measuring device, that is, a distance from the in-focus position. . The defocus amount may be acquired by other methods such as image processing.

  When the restored image is acquired, the position deviation amount acquisition unit 412 determines the position of the alignment mark 91 in the restored image by pattern matching between the restored image including the alignment mark 91 and the template image indicating the ideal alignment mark. Identified. Since the positional relationship between the imaging region by the imaging unit 2 and a reference position predetermined in the alignment apparatus 1, that is, the distance in the X direction and the Y direction between the imaging region and the reference position is known, the distance And the position shift amount from the reference position of the alignment mark 91 on the substrate 9 is obtained using the position of the alignment mark 91 in the restored image (step S43). At this time, the direction of the alignment mark 91 on the substrate 9 is also specified.

  The orientation of the alignment mark 91 is an inclination angle with respect to a predetermined reference direction on the XY plane. Usually, the substrate 9 is adjusted to some extent and positioned and placed on the stage 31. The inclination angle is very small.

  The control unit 42 controls the moving mechanism 3 based on the amount of misalignment and the orientation of the alignment mark 91, whereby the substrate 9 moves and rotates, and the alignment mark 91 is arranged at the reference position in a predetermined orientation. (Step S44). This makes it possible to perform various processes on the substrate 9 with high accuracy.

  Next, the effect when using a multi-valued image as the PSF will be described. As described above, conventionally, research on image restoration has been performed on natural images acquired using white illumination light. In this case, the PSF and the opening pattern are very similar, and the calculation can be performed by regarding the binary opening pattern as the PSF. In Non-Patent Document 1, when the aperture pattern is optimized, the aperture area is set to “1” and the light shielding area is set to “0”, and the PSF when the aperture stop is mounted is defined as the aperture pattern of the aperture stop. Assume that they are the same, and obtain an evaluation value.

  However, when the wavelength band of the illumination light is narrow, the PSF and the aperture pattern may differ greatly. At this time, the PSF is a multi-valued image that cannot be regarded as a binary image. For example, FIG. When a PSF is acquired with narrow-band light using the binary aperture pattern shown in FIG. A multi-valued image shown in B is acquired. FIG. In B, the point light source is arranged at a position slightly shifted from the in-focus position.

  9 is similar to FIG. A power spectrum 51 of the image of A, FIG. It is a figure which shows the power spectrum 52 of the image of B. These spectra are Y = 0, that is, on a straight line extending left and right through the center. As shown by the broken line 51, the aperture pattern is designed so that its power spectrum does not include a value close to 0. However, as shown by the solid line 52, the actual PSF includes a value close to 0. . As a result, an error in the deconvolution operation increases, and noise called ringing appears in the restored image. FIG. A shows an example of a blurred image, and FIG. B is FIG. A restored image obtained from A is shown. In the restored image, ringing appears around the black line. Ringing becomes noticeable when illumination light with a narrow wavelength band is used.

  FIG. A and FIG. B is a diagram showing another example in which PSF should not be expressed as a binary image in the case of light in a narrow wavelength band. FIG. A is a diagram showing a PSF obtained by imaging a white point light source using a circular aperture stop. FIG. B is a diagram showing a PSF obtained by imaging a point light source in a narrow wavelength band with a wavelength (620 ± 10) nm using a circular aperture stop. In the case of a white light source, a PSF with little unevenness is obtained, but in the case of light with a narrow wavelength range, a striped pattern called an Airy pattern appears in the PSF due to the influence of diffraction. Such an Airy pattern appears in the PSF obtained in step S22, and becomes a PSF that cannot be expressed by a binary image. In other words, the multi-gradation PSF is a PSF in consideration of the influence of diffraction. Note that the multi-tone PSF can also reflect the influence of an optical element such as a lens other than the influence of diffraction.

  Light in a narrow wavelength band is often used in industrial imaging devices in order to exhibit lens performance, and multi-tone PSFs are particularly suitable for industrial devices. In recent years, in a device manufacturing apparatus including a semiconductor circuit, a pattern is extremely fine, and a depth of field of an imaging unit is narrowed. Therefore, it is suitable for use in multi-tone PSF image restoration in an apparatus that requires extremely high precision positioning.

  When the circuit board is uneven, the inspection target has a certain depth, and there is a concern that the depth of field is insufficient. In this case, it is necessary to move the imaging unit in the direction toward the inspection object and perform imaging a plurality of times, or to perform imaging with a plurality of imaging units having different distances to the in-focus position. However, in the above embodiment, since the depth of field can be virtually expanded by image restoration, an appropriate image with reduced ringing can be acquired only by imaging once with one imaging unit 2. be able to.

  Since the technology using the coded aperture simply inserts the aperture stop 221 designed at the position of the pupil of the imaging unit (that is, the aperture position) as hardware, it can be easily introduced into various devices. it can.

  Next, an experimental example showing the effect of image restoration by the method of the present embodiment will be described. First, an opening pattern was designed by the method of the present embodiment. At this time, the PSF in the case where the point light source exists at a position shifted from the in-focus position by a defocus amount (+100) μm, that is, at a position 100 μm away from the in-focus position from the imaging unit was used. Further, PSF corresponding to the designed opening pattern was obtained by calculation at a pitch of 100 μm from the defocus amount (−300) μm to (+300) μm. Regarding Non-Patent Document 1, an aperture pattern was designed and a plurality of PSFs were calculated, and a plurality of PSFs were obtained in the case of a circular aperture.

  A circular aperture stop, an aperture stop according to the method of Non-Patent Document 1, and an aperture stop according to the present embodiment are attached to the imaging unit 2 in order, and the defocus amount (−300) μm to (+300) μm at a pitch of 100 μm, The alignment mark was imaged. The obtained image was restored using the corresponding PSF. Here, it is assumed that the defocus amount is known. The error of the detection position of the alignment mark from the restored image, that is, the Euclidean distance from the true value was evaluated.

  FIG. 12 is a diagram showing the evaluation results. The white diamond corresponds to a circular aperture stop, the white square corresponds to the aperture stop of Non-Patent Document 1, and the black circle corresponds to the aperture stop of the method of the present embodiment. Except for the case of the defocus amount (+300) μm, it can be seen that the position of the present embodiment can be detected with higher accuracy than the other methods. Further, the detection position error of 0.5 pixels or less is realized by the method of the present embodiment.

  FIG. 13 is a view showing a focused image of the alignment mark used for the evaluation. FIG. A is a view showing a blurred image with a defocus amount (+300) μm, that is, an out-of-focus image, acquired through a circular aperture stop. FIG. B considers the PSF as a binary circle and FIG. It is a figure which shows the image which restored the image of A. FIG. FIG. A is a diagram showing a blurred image with a defocus amount (+300) μm obtained through an aperture pattern obtained by the method of Non-Patent Document 1. FIG. FIG. B is a diagram illustrating an image restored by regarding PSF as being identical to a binary opening pattern. FIG. A is a diagram showing a blurred image with a defocus amount (+300) μm obtained through the opening pattern obtained by the method of the present embodiment. FIG. B is a diagram illustrating an image restored using PSF, which is a multi-valued image. FIG. B, FIG. B and FIG. As shown in FIG. It can be seen that the edge appears most clearly in B, ringing is small, and the image quality is good.

  FIG. A is a figure which shows the opening pattern designed by the method of this Embodiment in the said evaluation experiment, and is the same as that of FIG. FIG. B is a diagram showing a PSF obtained by actually placing a point light source at a position of a defocus amount (+100) μm. FIG. C is a diagram showing a PSF obtained by calculation. FIG. B and FIG. As shown in C, the actual PSF and the PSF simulated by computation are similar.

  18 is similar to FIG. A thru | or FIG. It is a figure which shows the power spectrum on the straight line extended to Y = 0 of the image shown to C, ie, right and left through a center. The broken line 53 is shown in FIG. A solid line 54 corresponds to FIG. Corresponding to B, the solid line 55 is shown in FIG. Corresponds to C. Also in FIG. 18, the PSF spectrum obtained as a multivalued image by calculation is closer to the actual PSF spectrum than the aperture pattern spectrum. Unlike the case illustrated in FIG. 9, the spectrum of the actual PSF and the PSF of the multi-value image does not include a value close to 0.

  As described above, the image restoration according to the present embodiment can realize a good image restoration even when the wavelength band of light from the light source is narrow.

  In the design of the opening pattern, a plurality of images each indicating a plurality of alignment marks are prepared, and a plurality of power spectra in the plurality of images are obtained for each direction. Then, the representative spectrum is acquired from the plurality of power spectra. A plurality of candidate patterns are prepared, and an evaluation value indicating the restoration accuracy of an image captured using each candidate pattern is obtained for each direction using the noise level and the representative spectrum in the imaging unit 2. . Thereafter, by obtaining final evaluation values from evaluation values in a plurality of directions for each candidate pattern, an opening pattern used in the imaging unit 2 is determined from the plurality of candidate patterns.

  Thereby, an opening pattern specialized for the alignment mark can be determined. As a result, the alignment device 1 in which the opening pattern 221 is provided in the imaging unit 2 can accurately restore an image as an image restoration device, and the alignment mark 91 on the substrate 9 is accurately arranged at the reference position. It becomes possible.

  In designing the aperture pattern, the aperture pattern can be determined more appropriately by obtaining the evaluation value using the noise level measured for the image sensor 23 of the imaging unit 2.

  The imaging target in the opening pattern design may be other than the alignment mark, and the target on which the imaging target is formed may be other than the substrate. In the case of an imaging target of another category, in the process relating to the design of the opening pattern in FIG. 3, a plurality of images of the other category are prepared in step S11. The other processes are the same as described above, and thereby, an opening pattern specialized for the other category can be determined. As described above, in the process related to the design of the opening pattern, by preparing a plurality of images respectively indicating a plurality of imaging objects belonging to the same category in step S11, determining an appropriate opening pattern according to the category. Can do. As a result, in the image restoration apparatus including the imaging unit 2 and the image restoration unit 411, a restored image can be accurately acquired from an image obtained by imaging an imaging target belonging to the category.

  Of course, the above-described image restoration using a multi-tone PSF may be used for an imaging target whose category is not specified.

  The image restoration is suitable for industrial devices that use illumination light in a narrow wavelength band. Therefore, in many cases, the imaging target is an artificial object, and the target image shows a substantially binary pattern in which a bright region and a dark region exist.

  In the above embodiment, the wavelength range of the light emitted from the point light source or the illumination unit 21 may be so narrow that it can be regarded as a single wavelength. The wavelength range is so narrow that an Airy pattern appears in the PSF, and the half width is preferably 40 nm or less, more preferably 20 nm or less. The Airy pattern only needs to be recognized so that it does not need to appear clearly.

  In the above embodiment, the aperture stop 221 having an aperture pattern corresponding to one defocus amount is also used for restoring a blurred image obtained with another defocus amount. In this case, the imaging target is not necessarily located at a predetermined position where it is assumed that the point light source is arranged when the opening pattern is determined. In the alignment apparatus 1, the imaging target is arranged at the predetermined position or in the vicinity thereof.

  On the other hand, an opening pattern may be prepared for each defocus amount. In this case, a plurality of aperture stops 221 may be switched, and a variable aperture stop that is a liquid crystal array may be provided in the imaging unit 2. When a plurality of opening patterns are used, the defocus amount may be acquired at the time of imaging. The defocus amount may be measured, estimated from the acquired image, or obtained from a plurality of image restoration results.

  The evaluation value for each candidate pattern may be obtained by using an expression obtained by appropriately correcting Expression 2 or an expression different from Expression 2. The representative spectrum may be acquired from one image indicating the imaging target belonging to one category. In this case, the power spectrum in each direction in the one image is acquired as the representative spectrum. In the calculation of the evaluation value, representative spectra in all directions are obtained, but only representative spectra in some directions may be used for calculation of evaluation values.

  Other evaluation values may be used. For example, the minimum value of the power spectrum of PSF may be used as the evaluation value. In this case, the higher the minimum value, the higher the evaluation. In general terms, this indicates accuracy when a blurred image obtained when an imaging target is placed at a predetermined position using each of the plurality of point spread functions obtained in step S22. If it is a thing, various evaluation values can be used.

  As long as one opening pattern is selected from a plurality of candidate patterns based on the evaluation value, various methods may be employed. In the above embodiment, selection of candidate patterns and calculation of evaluation values are performed substantially simultaneously. For example, a large number of candidate patterns are simply created at random, and evaluation values for each candidate pattern are obtained. One candidate pattern may be selected based on.

  In the above embodiment, the PSF corresponding to the candidate pattern is efficiently obtained by calculation. However, a large number of aperture stops 221 having the candidate pattern may be actually created, and the PSF may be acquired using a point light source. The aperture stop 221 having a coded opening may be cut out from paper, a metal plate, or the like, and a liquid crystal shutter may be used as the stop.

  In the above embodiment, the alignment mark is imaged using reflected light from the substrate. However, in the case of a glass substrate or the like, the alignment mark may be imaged using transmitted light from the substrate. The alignment apparatus 1 can be used for various apparatuses other than the inspection apparatus and the drawing apparatus. For example, it can be used for a high-precision printing apparatus. Further, the image restoration apparatus including the imaging unit 2 and the image restoration unit 411 may be provided in another apparatus (for example, an inspection apparatus) that acquires an image in addition to the alignment apparatus 1.

  The configurations in the above-described embodiments and modifications may be combined as appropriate as long as they do not contradict each other.

1 Positioning device (image restoration device)
2 Imaging unit 21 Illumination unit 91 Alignment mark (imaging target)
221 Aperture stop 411 Image restoration unit 412 Position shift amount acquisition unit S11 to S18, S21 to S30, S41 to S44 Steps

Claims (8)

A method of designing an aperture pattern used in an imaging unit of an image restoration device,
a) preparing a plurality of candidate patterns which are opening patterns;
b) When a point light source is arranged at a predetermined position, a point spread function that is acquired by the imaging unit via each of the plurality of candidate patterns and in which an Airy pattern appears is expressed as a multi-tone Acquiring as an image;
c) Evaluation value indicating accuracy when a blurred image obtained when the imaging target is placed at the predetermined position is restored using each of the plurality of point spread functions obtained in the step b). A process of obtaining
d) selecting one of the plurality of candidate patterns based on the plurality of evaluation values obtained in the step c);
A method for designing an opening pattern, comprising: A method for designing an opening pattern according to claim 1,
In the step b), the point spread function is acquired by performing ray tracing on light emitted from a virtual point light source. An opening pattern design method according to claim 1 or 2,
A method for designing an aperture pattern, wherein the evaluation value obtained in the step c) is obtained using a representative spectrum obtained from a spectrum of a plurality of images indicating a plurality of imaging objects belonging to one category. A method for designing an opening pattern according to claim 3,
When the frequency is ξ, the representative spectrum is S _ξ , the Fourier transform of the PSF corresponding to the candidate pattern is K _ξ , and the noise level is σ, the evaluation value R of the candidate pattern is
A method for designing an opening pattern, characterized in that An image restoration method,
The same wavelength as the light emitted from the point light source or the same wavelength by the imaging unit provided with an aperture stop having an aperture pattern determined by the aperture pattern design method according to claim 1. Capturing an image of an imaging target irradiated with a band of light, and obtaining an image;
Using the multi-tone point spread function corresponding to the opening pattern to obtain a restored image from the image;
An image restoration method comprising: An image restoration device,
An imaging unit provided with an aperture stop having an aperture pattern determined by the aperture pattern design method according to any one of claims 1 to 4,
An illumination unit that irradiates the imaging target with light having the same wavelength as the light emitted from the point light source
An image restoration unit that obtains a restored image using a multi-tone point spread function corresponding to the aperture pattern from an image obtained by imaging the imaging target by the imaging unit;
An image restoration apparatus comprising: The image restoration device according to claim 6,
The image restoration device, wherein the illumination unit includes a laser, a light emitting diode, or a combination of a white light source and a narrow band-pass filter as a light source. The image restoration device according to claim 6 or 7,
An image restoration apparatus, further comprising: a positional deviation amount acquisition unit that obtains a positional deviation amount from a reference position of the imaging target by pattern matching between the restored image of the imaging target that is an alignment mark and a template image. .