JP4235756B2 - Misalignment detection method, misalignment detection apparatus, image processing method, image processing apparatus, and inspection apparatus using the same - Google Patents

Description

[0001]
BACKGROUND OF THE INVENTION
The present invention relates to a misregistration detection method and misregistration detection apparatus that detect misregistration between images, and more particularly to an image processing method, an image processing apparatus, and an image processing method for accurately correcting misregistration between images in a short time. The present invention relates to an inspection apparatus and an inspection method using the same.
[0002]
[Prior art]
A CCD sensor is used in a conventional semiconductor device, liquid crystal display device, photomask or reticle pattern shape defect inspection device, and the presence or absence of a defect is detected from the image. In such a defect inspection apparatus, a die-to-die method (Die-to-Die) that compares adjacent identical pattern shapes and a die-to-database method (Die-to-Database) that generates a reference image based on CAD data and compares it with it. There is. (For example, see Patent Document 1)
[0003]
In the die-to-die method, two identically shaped die patterns are detected by different CCD sensors. Alternatively, after detecting the die pattern once by one CCD sensor, the same die pattern is detected by moving the substrate or the CCD sensor. By comparing the two digital images, the presence or absence of a defect is determined. When comparison processing is performed on two digital images, each pixel value may be subtracted between the two digital images. That is, in a pixel having a different value, it becomes a value other than 0, and a defect has occurred in this place. In this way, slight differences between two images having the same content (ie, pattern shape defects) can be found. Some image data are subjected to differentiation processing and then subjected to two-dimensional FFT processing to identify an identification object. (For example, Patent Document 2) In this method, an identification object is imaged by a CCD camera, and differentiation processing is performed on the image data. Then, two-dimensional FFT processing is performed, and the image is identified by the peak value of the spectrum, that is, the maximum value (or extreme value) of the coefficient of the frequency expressed by the FFT processing.
[0004]
However, defect detection by these methods is based on the premise that there is no shift between images. If there is a slight shift between images, even if the same image is subtracted, the shift is erroneously detected as a pseudo defect. Therefore, an inspection apparatus that performs defect inspection by this method must perform inspection while suppressing the cause of deviation as much as possible. Therefore, it is necessary to accurately match the CCD sensor to the same position in different patterns.
[0005]
The configuration of this die-to-die type defect inspection apparatus will be described with reference to FIG. Reference numeral 1 denotes a CCD linear image sensor, 2 a substrate, and 3 an image processing apparatus. Various patterns are formed on the substrate 2 to be inspected, and two CCD linear image sensors 1 are provided facing the substrate 2. A two-dimensional image is detected by the two CCD linear image sensors 1. Then, by moving the substrate 2 in the direction of the arrow at a constant speed, two-dimensional data of the same die pattern image is detected. These two images are taken into the image processing device 3. And the difference of the two-dimensional data between images was calculated | required, and the presence or absence of the defect was discriminate | determined by the difference. Note that an actual defect inspection apparatus may be provided with an optical system such as a light source for illuminating the substrate 2 and a lens and a mirror for guiding light from the substrate to the CCD linear image sensor 1.
[0006]
In order to inspect the entire surface of the substrate, either one or both of the substrate 2 and the CCD linear image sensor 1 are moved. An example of a configuration for performing the defect inspection on the entire surface of the substrate will be described below. The substrate 2 is continuously moved in the direction of the arrow in FIG. 8, and the two-dimensional data of the image is captured by the CCD linear image sensor 1 at regular time intervals. When the measurement of the continuously moved line is completed, the CCD linear image sensor 1 is moved in parallel with the substrate perpendicular to the arrow. When the movement is completed, the substrate 2 is continuously moved again in the direction of the arrow. By repeating this operation, the entire substrate is inspected for defects.
[0007]
The positional deviation correction between the two images is performed by adjusting the position and inclination of the CCD linear image sensor 1 or the substrate 2. Alternatively, adjustment is performed by a mirror or a lens provided in the middle of the optical path to the CCD linear image sensor 1. However, it is very difficult to perform mechanically or optically, and pseudo defects may occur. Further, when a positional deviation occurs after the CCD linear image sensor 1 or the substrate 2 is moved, the position must be adjusted each time.
[0008]
There is also an image processing method that corrects misalignment on an image, and there is a method that can correct up to sub-pixels. However, recently, the width of the wiring pattern has become narrower. Therefore, it is necessary to increase the resolution (detection range around one pixel) of the CCD linear image sensor in order to perform inspection with higher accuracy. Furthermore, the requirements for the accuracy of this misregistration correction are becoming stricter. Therefore, when a wide range of images is acquired with the CCD sensor resolution increased with a conventional inspection apparatus, there is a problem that the amount of image data increases and the required calculation time and storage capacity increase. When the calculation time increases, there is also a problem that the inspection throughput is slowed down.
[0009]
[Patent Document 1]
U.S. Pat. No. 4,805,123
[Patent Document 2]
JP 2000-357236 A
[Patent Document 3]
JP-A-11-201908
[0010]
[Problems to be solved by the invention]
As described above, in the conventional inspection apparatus, when the two-dimensional images are subjected to the positional deviation correction and the comparison processing is performed, there is a problem that a necessary calculation time and a storage capacity are increased.
[0011]
The present invention has been made in view of such problems, and an object of the present invention is to provide a displacement detection method and a displacement detection apparatus that can detect displacement between two images at high speed and with high accuracy. To do. It is another object of the present invention to provide an image processing method, an image processing apparatus, an inspection apparatus, and an inspection method using them.
[0012]
[Means for Solving the Problems]
A positional deviation detection method according to the present invention is a positional deviation detection method for detecting a positional deviation amount between first image data and second image data, and (a) 1 obtained from the image data. A step of Fourier transforming the dimension data (for example, step S6 according to the embodiment of the present invention), and (b) a step of obtaining a phase at at least one frequency based on the result of step (a) (for example, Step S8) according to the embodiment of the present invention, and (c) the amount of misalignment between images based on the difference between the phase obtained in step (b) and the phase of the second image data at the frequency. Is obtained (for example, step S9 according to the embodiment of the present invention). This makes it possible to accurately detect the amount of misalignment between images.
[0013]
In the positional deviation detection method according to the present invention, in the step (a) of the positional deviation detection method described above, coefficients corresponding to a plurality of frequencies are obtained by the Fourier transform, and in the step (b), the plurality of frequencies are detected. The intensity at each frequency is obtained, the frequency is extracted based on the extreme value of the intensity, and the phase is obtained at the extracted frequency. This makes it possible to accurately detect the amount of misalignment between images.
[0014]
In the above-described misregistration detection method, further, a step of acquiring two-dimensional data based on the first image data (for example, step S2 according to an embodiment of the present invention), and based on the two-dimensional data, the 1 You may have the step (For example, step S4 concerning embodiment of this invention) which produces | generates projection waveform data as dimension data. This makes it possible to accurately detect the amount of misalignment between images.
[0015]
In the positional deviation detection method according to the present invention, the step of generating the projection waveform in the positional deviation detection method described above generates one-dimensional projection waveform data for a plurality of different directions, and the one-dimensional projection waveform data. Steps (a), (b) and (c) are executed. Thereby, it is possible to accurately detect the amount of displacement in two directions.
[0016]
An image processing method according to the present invention is an image processing method for performing a comparison process between first image data and second image data, and a step of obtaining a displacement amount of any of the displacement detection methods described above. After that, a step of correcting the positional deviation amount between the images with respect to the image data of the first image data or the second image data (for example, step S10 according to the embodiment of the present invention), The method includes a step (for example, step S11 according to the embodiment of the present invention) for comparing the corrected image data. As a result, it is possible to accurately correct misalignment between images.
[0017]
An inspection method according to the present invention is an inspection method for performing an inspection based on a comparison processing result between first image data and second image data, the step of comparing the image data of the above-described image processing method. Thereafter, the method includes a step of outputting an inspection signal based on the comparison result subjected to the comparison processing (for example, step S12 according to the embodiment of the present invention). As a result, it is possible to compare images that have been accurately corrected for misalignment, thereby improving the inspection accuracy.
[0018]
A misregistration detection apparatus according to the present invention is a misregistration detection apparatus that detects a misregistration amount between first image data and second image data, and is a one-dimensional image obtained from the first image data. First computing means for Fourier transforming data; second computing means for obtaining a phase at at least one frequency based on a result of the first computing; and the first computing means obtained in the second computing. Third computing means for obtaining a positional deviation amount between the first image data and the second image data based on a difference between the phase of the image data and the phase of the second image data at the frequency. (For example, DSP33 concerning embodiment of this invention) is provided. As a result, the amount of misalignment between images can be detected accurately in a short time.
[0019]
The positional deviation detection device according to the present invention obtains coefficients corresponding to a plurality of frequencies by the Fourier transform in the first calculation of the above-described positional deviation detection method, and in the second calculation, The intensity at each frequency is obtained, the frequency is extracted based on the extreme value of the intensity, and the phase is obtained at the extracted frequency. As a result, the amount of misalignment between the image data can be accurately detected in a short time.
[0020]
In the above-described positional deviation detection apparatus, the first image data is two-dimensional data, and a projection waveform generation circuit that generates projection waveform data as one-dimensional data based on the two-dimensional data (for example, an embodiment of the present invention) Further, a projection waveform generation circuit 31) may be included. As a result, the amount of misalignment between images can be detected accurately in a short time.
[0021]
In the above-described positional deviation detection apparatus, the projection waveform generation circuit generates one-dimensional projection waveform data for a plurality of different directions, and the first calculation, the second calculation, and the one-dimensional projection waveform data. The third calculation may be executed. As a result, the amount of displacement in the two directions can be detected accurately in a short time.
[0022]
The projection waveform generation circuit of the above-described positional deviation detection apparatus may generate the projection waveform by integrating the two-dimensional data in the first direction. As a result, the amount of misalignment between images can be detected accurately in a short time.
[0023]
The positional deviation detection apparatus according to the present invention differentiates the two-dimensional data in a second direction orthogonal to the first direction before generating the projection waveform in the projection waveform generation circuit of the positional deviation detection apparatus described above. Differential processing is performed. Thereby, edge detection can be performed, and the amount of positional deviation between images can be detected accurately in a short time.
[0024]
An image processing apparatus according to the present invention is an image processing apparatus that performs a comparison process between first image data and second image data, and the misregistration detection apparatus according to any one of claims 7 to 12, A shift correction circuit (for example, the shift correction circuit 34 according to the embodiment of the present invention) that corrects the positional shift amount with respect to the original image, and an image comparison circuit (for example, this book) that compares the corrected image. It has an image comparison circuit 36) according to an embodiment of the invention. As a result, the amount of misalignment between images can be corrected accurately in a short time.
[0025]
An inspection apparatus according to the present invention is an inspection apparatus that performs an inspection based on a comparison processing result between first image data and second image data, and includes an imaging unit that captures an image of an object to be inspected (for example, Based on the result of the first calculation, the CCD linear image sensor 1) in the embodiment of the present invention, the first calculation means for Fourier transforming the one-dimensional data obtained from the first image data, and at least Based on a second computing means for obtaining a phase at one frequency, a phase of the first image data obtained in the second computation, and a phase of the second image data at the frequency, Calculation means (for example, DSP 33 in the embodiment of the present invention) for performing a third calculation for obtaining the amount of positional deviation between the first image data and the second image data; A displacement correction circuit (for example, the displacement correction circuit 34 in the embodiment of the present invention) that corrects the positional deviation amount obtained by the amount detection device with respect to the first image data or the second image data; An image comparison circuit that compares the corrected image data (for example, the image comparison circuit 36 in the embodiment of the present invention) and an inspection signal based on the comparison result are output. As a result, it is possible to compare images that have been accurately corrected for misalignment, thereby improving the inspection accuracy.
[0026]
In the above-described inspection apparatus, it is preferable that the first image data is two-dimensional data, and further includes a projection waveform generation circuit that generates projection waveform data as one-dimensional data based on the two-dimensional data. As a result, it is possible to compare images that have been accurately corrected for positional deviation in a short time, so that the inspection accuracy can be improved.
[0027]
In the above-described inspection apparatus, it is desirable to further include storage means for storing the amount of positional deviation. Thereby, the convenience of an inspection apparatus can be improved.
[0028]
DETAILED DESCRIPTION OF THE INVENTION
Embodiment 1 of the Invention
A configuration of an inspection apparatus using the positional deviation detection method and the image processing method according to the present invention will be described with reference to FIG. FIG. 1 is a perspective view showing a configuration of an inspection apparatus. Reference numeral 1 denotes a CCD linear image sensor, 2 a substrate, and 3 an image processing apparatus.
[0029]
Two CCD linear image sensors 1 are provided opposite to the substrate 2 to be inspected. The CCD linear image sensor 1 has 1024 effective pixels and can acquire one-dimensional line data. By using the CCD linear image sensor 1, image distortion can be suppressed and inspection can be performed with high accuracy. A digital signal of 8 bits, that is, 0 to 255 is accumulated in one pixel. Further, by adjusting the lens system of the CCD linear image sensor 1, the spatial resolution of the CCD linear image sensor 1 is about 0.125 μm / pixel.
[0030]
Although an optical system is not particularly illustrated in FIG. 1, optical components such as a light source, a lens, a mirror, a half mirror, and a beam splitter may be provided on the optical path in terms of the device configuration. The substrate 2 is supported by a holder (not shown), and this holder is moved at a constant speed in the direction of the arrow. As a result, a two-dimensional image relating to the substrate 2 is captured. The line data for 1024 pixels per line is captured 512 times by the movement of the substrate 2. Since the data of 1024 × 512 pixels is processed at a time, it becomes two-dimensional image data. Such two-dimensional data is acquired at a cycle of 15 msec. And said process is performed continuously.
[0031]
An example of an image acquired by the CCD linear image sensor 1 is shown in FIG. 4 is an image and 5 is a wiring. For this explanation, these two images 4 are denoted as L and R. The one-dimensional CCD linear image sensor 1 acquires line data in the Y direction, and the substrate 2 is moved in the X direction. That is, digital signals for 512 pixels in the X direction and 1024 pixels in the Y direction are acquired. In the L and R images 4, an image of a portion where the wiring 5 on the substrate is formed is acquired. However, since the CCD linear image sensor 1 captures the position where the same pattern is shifted, the wiring 5 is also shifted.
[0032]
An inspection method for performing an inspection by correcting a positional deviation between images will be described with reference to FIGS. FIG. 3 is a flowchart of an inspection method for performing inspection by performing misalignment correction, and FIG. 4 is a block diagram showing a configuration of an image processing apparatus using misalignment correction. Here, 31 is a projection waveform generation circuit, 32 is a FIFO memory, 33 is a DSP (Digital Signal Processor), 34 is a shift correction circuit, 35 is an image memory, and 36 is an image comparison circuit. These are provided inside the image processing apparatus 3 of FIG.
[0033]
First, as described above, two one-dimensional images are acquired by the two CCD linear image sensors 1 (step S1). A two-dimensional image is generated by moving the substrate 2 (step S2). The two two-dimensional images are synchronized with the horizontal synchronization signal and the vertical synchronization signal and input to the projection waveform generation circuit 31. In the projection waveform generation circuit 31, the two-dimensional data is differentiated in the X direction (step S3). Subsequently, the differentiated two-dimensional data is integrated in the Y direction (step S4). Thereby, one-dimensional projection waveform data is generated from the two-dimensional data.
[0034]
These two projection waveforms are stored in the FIFO memory 32 (step S5). The DSP 33 takes in the waveform data from the FIFO memory 32 and performs a fast Fourier transform (hereinafter, FFT) process (step S6). As a result, the waveform data that is a function of position is converted into a function of frequency. A frequency in the vicinity of the extreme value of the intensity obtained by the FFT process is extracted (step 7). A phase is calculated at the frequency (step S8). Then, the phase difference between the two images is obtained at the frequency, and the shift amount is detected from the phase difference (step S9).
[0035]
The DSP 33 sets the shift amount between the two images in the shift correction circuit 34, and performs shift correction on the two images to be compared (step S10). The two images to be compared are stored in the image memory 35 at the same time, and the shift correction is performed by the shift amount obtained in step 9. The image comparison circuit 36 performs image comparison processing (step S11) using the contents of the image memory 35 subjected to the shift correction. Here, the difference of the two-dimensional data between both images in which the shift amount is corrected is taken. And the presence or absence of a defect is judged from this difference (step S12). That is, the value of the difference is not 0 at a place where there is a difference between both images. Therefore, it is determined that there is a defect at the inspection location of the substrate 2, and a defect detection signal is output (step S13). When it is determined that there is no defect, a signal indicating no defect may be output or the process may be terminated as it is. The image between the two CCD linear image sensors corrected for positional deviation in the process as described above is subjected to a comparison process, and defect inspection can be performed. Either one or both of the CCD linear image sensor 1 and the substrate 2 are moved, the above processing is performed, and the entire surface of the substrate 2 is inspected for defects. Further, when optical components such as a lens and a mirror are provided between the CCD linear image sensor 1 and the substrate 2, the entire substrate may be inspected for defects by adjusting them.
[0036]
Next, signal processing in steps S2 to S4 will be described in detail. Assume that the image shown in FIG. 2 is generated in step S2. This image is assumed to have an effective pixel area of 1024 pixels in the Y direction and 512 pixels in the X direction. And one wiring 5 is imaged for each. However, there is a positional shift between images, and the position of the wiring 5 on the image is shifted. Different values are acquired by the CCD linear image sensor 1 in the area where the wiring 5 is provided and in the area where the wiring 5 is not provided.
[0037]
The waveform of the line data in the direction of the arrow shown in FIG. 2 is shown in FIG. This line data is assumed to be the line data of the uppermost pixel in the Y direction. In the image of L, the position where the region where the wiring 5 is provided is changed to the region where the wiring 5 is not provided is represented by x. ₁ And Similarly, in the L image 4, the position where the region where the wiring 5 is provided is changed to the region where the wiring 5 is provided is represented by x. ₂ And In the R image 4, it is assumed that the L image is shifted by Δx. In addition, since the same pattern shape is imaged in the L and R images 4, the thickness of the wiring 5 is the same. Further, it is assumed that the same value is acquired in the region where the wiring 5 is provided. Similarly, it is assumed that a value of 0 is acquired for all areas where no wiring is provided.
[0038]
Accordingly, as shown in FIG. 5A, the line data of the image 4 of L is x in which the wiring 5 is provided. ₁ ~ X ₂ It becomes a rectangular shape protruding only the area of. Similarly, the line data of the R image 4 is x ₁ + Δx to x ₂ Only a region of + Δx protrudes into a rectangular shape. When differentiation processing (difference processing) is performed on the waveform data, a spectral component appears only at a position where the value changes. Therefore, the waveform is as shown in FIG. Thereby, the edge detection of the wiring 5 can be performed.
[0039]
Further, the above process is performed for all lines in the Y direction. Then, the waveform as shown in FIG. 5B is obtained in all lines. Next, integration processing is performed on the data for all the lines in the Y direction. That is, the sum is obtained for data having the same position in the X direction. As a result, the projection waveform in which the edge portion of the wiring 5 is emphasized as shown in FIG. The above differentiation processing and integration processing are performed by the projection waveform generation circuit 31. As a result, the two-dimensional image data can be converted into one-dimensional projection waveform data, and the calculation time required for the shift correction detection including the subsequent FFT processing can be reduced. For example, when a two-dimensional image of 1024 × 512 pixels is converted into a one-dimensional projection waveform, the shift amount can be obtained in about 1 msec. Shortening the calculation time leads to shortening of the inspection time, and it is also possible to improve the throughput of the inspection apparatus. The projected waveform is one-dimensional data generated based on the two-dimensional data, and can be obtained by integrating the two-dimensional data in one direction as described above. Of course, the differentiation process may not be performed. Further, the integration process may be performed only on a specific part. In addition to the integration process, a projection waveform can be obtained by extracting one line data from the two-dimensional data. In this case, it is desirable to extract the data of the line including the edge after performing the differentiation process.
[0040]
Next, frequency extraction in step S7 and phase calculation in step S8 will be described in detail with reference to FIG. FIG. 6 is an example of a power spectrum after the projection waveform data is subjected to FFT processing. The one-dimensional waveform is a finite digital signal, and Fourier coefficients corresponding to a plurality of frequencies can be obtained by discrete Fourier transform (DFT), and the amplitude (intensity) and phase can be calculated. That is, the amplitude (intensity) and phase at the frequency are calculated from the real part and imaginary part of each frequency component, and the power spectrum and phase spectrum can be obtained. Thus, amplitude (intensity) and phase can be expressed as a function of frequency. The positional deviation amount can be obtained from this phase. In order to speed up the processing, it is desirable to use FFT for these Fourier transforms. Since the image data is converted into one-dimensional projection waveform data, it is possible to perform arithmetic processing at a higher speed than two-dimensional FFT. As a result, the calculation time can be shortened, it is not necessary to use a high-performance image processing apparatus, and the cost can be reduced.
[0041]
An example of the power spectrum obtained by this FFT processing is shown in FIG. FIG. 6 shows the intensity (square of amplitude) obtained by performing FFT processing on the projection waveform of the image data of the CCD linear image sensor 1 as a function of frequency. It is not related to the images shown in FIGS. First, an extreme value (peak) having a component of level B or higher is searched for at a frequency of A or higher. Here, two peaks, peak 1 and peak 2, appear. The frequency at peak 1 and peak 2 is f ₁ , F ₂ And Consider peak 1 first. Frequency f ₁ Obtain the phase of the L and R images in the phase difference Δθ ₁ And Note that the phase and phase difference may be calculated by calculating only the phase and phase difference at that frequency, or after obtaining the phase spectrum and phase difference spectrum as a function of frequency. Where the phase difference between the two images is Δθ ₁ , Its angular velocity is ω ₁ (= 2πf ₁ ), Shift amount (pixel) = Δθ ₁ / Ω ₁ (However, ω ₁ = 2πf ₁ = 2πn / N, where N is the total number of pixels in the waveform and n is 1 to N / 2). Also, the frequency f at peak 2 ₂ The same processing is performed for the phase difference Δθ ₂ To calculate the amount of deviation. This average is the amount of shift between images. Note that the shift amount may be calculated by weighting the peak value. This deviation amount enables accurate and highly accurate correction. It should be noted that the frequency f from either power spectrum of L or R ₁ May be extracted to determine the amount of displacement. Further, two frequencies may be extracted from the extreme values of the L and R power spectra, and an average value obtained from the two phase differences may be used as the positional deviation amount. Since the edge frequency of the image is included in the extreme value frequency of the power spectrum, it is possible to perform accurate and highly accurate misalignment detection and misalignment correction, and to improve inspection accuracy. Thereby, the amount of positional deviation can be obtained with high accuracy. In addition, by searching for peaks having components above a certain level, the number of frequencies to be extracted can be reduced, leading to a reduction in calculation time. Further, by searching for a frequency higher than a certain frequency, low frequency components such as stage vibrations are not searched, and a more accurate inspection can be performed. Further, the target peaks are not limited to two and may be one or more. Furthermore, by performing this calculation process on all the images to be compared, even when the optical system is shifted with time, it is possible to cope with a change in the amount of positional shift. Therefore, a more accurate inspection can be performed.
[0042]
Embodiment 2 of the Invention
The present embodiment is different from the inspection method shown in the first embodiment in the frequency extraction method. Accordingly, the steps other than step S7 are the same, and the configurations of the inspection apparatus and the image processing apparatus are also the same, and the description thereof is omitted.
[0043]
In the first embodiment, an extreme value, that is, a phase difference at a peak is obtained and used as a shift amount between images. In the present embodiment, as shown in FIG. 7, the frequency at half the intensity of peak 1 (extreme value) is f. ₃ , F ₄ And F ₃ And f ₄ Is a frequency that is half the intensity at the frequency closest to the frequency at the extreme value. Similarly, the frequency at half the intensity of peak 2 (extreme value) is f ₅ , F ₆ And And f ₃ ~ F ₄ And f ₅ ~ F ₆ The phase of L and R is calculated | required in each frequency of this frequency band, and the phase difference in each frequency is calculated | required. A shift amount (Δθ / ω) is obtained from each phase difference Δθ, and the average value is used as a positional shift amount between images. By calculating the phase difference from the frequency band having a value close to the extreme value and calculating the shift amount, an accurate positional shift amount can be determined in a short time.
[0044]
Note that the amount of deviation (pixels) between images is Δθ / ω as in the first embodiment. Deviation correction is performed by using an average value of deviation amounts obtained from a plurality of extracted frequencies as a positional deviation amount. Thereby, even when the peak of the power spectrum is gentle or when there is a peak of the same intensity in the vicinity of the maximum extreme value, the deviation amount can be obtained accurately. The position of the frequency for obtaining the phase difference is not limited to a position that is 1/2 of the peak value, and any value that is greater than 0 and less than 1 may be used. One or more frequencies may be extracted. Since the edge information of the image is included in the frequency in the vicinity of the extreme value, it is possible to perform accurate and highly accurate misalignment detection and misalignment correction, and to improve the inspection accuracy. In addition, since the number of frequencies to be extracted can be reduced, the calculation time is shortened.
[0045]
Embodiment 3 of the Invention
In the present embodiment, the method of obtaining the projection waveform in the image processing method shown in the above-described embodiment is different. Therefore, steps other than step S3 to step S5 are the same, and the configurations of the inspection apparatus and the image processing apparatus are also the same, and the description thereof is omitted.
[0046]
The projection waveform in the first embodiment has been subjected to integration processing in the Y direction after performing differentiation processing in the X direction. The present embodiment is different in that a differential process is further performed in the Y direction, an integration process is then performed in the X direction, and two projection waveforms are generated. An FFT process is performed on each of the projection waveforms in such two directions to obtain a phase difference. By correcting the amount of deviation in two directions from this phase difference, more accurate two-dimensional correction can be performed. Note that, after the deviation amount is obtained from the projection waveform in the X direction and the deviation correction is performed, the deviation amount may be obtained again from the projection waveform in the Y direction and the deviation correction may be performed. Conversely, the deviation amount in the X direction may be corrected after obtaining the deviation amount from the projected waveform in the Y direction and correcting the deviation. As a result, accurate displacement correction in two directions can be performed in a shorter time.
[0047]
Other embodiments.
The circuit that performs the FFT processing shown in the above-described embodiment is not limited to the DSP 33, and it is only necessary to be able to perform arithmetic processing for obtaining the shift amount from the FFT processing and the phase difference. These processes may be performed by different circuits. In addition, since the DSP 33 can perform calculations with sub-pixel accuracy, more accurate and accurate correction can be performed. You may further provide the memory which memorize | stores the deviation | shift amount calculated by DSP33. For example, the positional deviation amount is obtained only for the first image, and the positional deviation amount is stored in the deviation correction circuit 4. This deviation amount is applied to a certain number of images. Then, after obtaining a certain number of images and performing comparison processing, the deviation amount is updated. More specifically, for example, after the inspection of one substrate is completed and the next substrate is set, the displacement amount is updated and the inspection is performed. Thereby, the arithmetic processing performed each time can be omitted. Furthermore, when the amount of deviation cannot be obtained because of poor image processing, the stored amount of deviation may be used. Even when the amount of deviation is detected for each image capture, there may be no pattern on the substrate and the amount of deviation may not be obtained. In this case, if the deviation amount obtained in the past is referred to, it is possible to cope with the case where the deviation amount cannot be obtained. For example, when an image without a pattern is taken on a substrate and foreign matter is attached to one of the images, the amount of deviation may not be obtained, which is effective. Furthermore, since the distribution of the deviation amount in the substrate surface represents the moving characteristics of the stage or CCD sensor on which the substrate is placed, the previous deviation amount can be referred to even when the substrate changes. For example, when the substrate or the CCD sensor is tilted and moved, it is possible to obtain the tilt and correct the positional deviation based on the movement characteristics.
[0048]
In the above-described embodiment, the two-dimensional image data is acquired by moving the substrate 2 using the one-dimensional CCD linear image sensor 1, but the CCD linear image sensor 1 may be moved. Of course, a two-dimensional CCD sensor may be used. Further, one-dimensional image data comparison processing may be performed. In this case, steps S2 to S4 for generating projection waveform data and the projection waveform generation circuit 31 are not necessary. The image processing method according to the present invention is suitable for a die-to-die inspection apparatus using two CCD sensors. Not only the die-to-die method but also a die-to-database type inspection apparatus using one CCD sensor can be used. Further, a plurality of image data may be acquired by one CCD linear image sensor and the comparison process may be performed. Three or more CCD linear image sensors may be used. The detector is not limited to the CCD sensor, and other detectors may be used. For example, an individual imaging device such as a MOS (Metal Oxide Semiconductor) or a CMOS (Complementary MOS) may be used.
[0049]
The number of pixels, the image, the spectrum, and the like shown in the above embodiment are examples, and are not limited to those shown in the drawings. The above-described image processing method and image processing apparatus are preferably used for an inspection apparatus that performs a comparison process between two images. Further, it may be used only for detecting the amount of displacement. Note that the inspection object may be placed vertically, horizontally, or obliquely. Further, it is preferable to use the inspection object as a semiconductor device, a liquid crystal display device, a photomask, or a reticle for a defect inspection apparatus that performs defect inspection of these patterns. The semiconductor device may inspect a plurality of semiconductor chips after dicing. Further, a plurality of inspection objects having the same shape may be inspected, or a plurality of inspection objects having the same pattern may be inspected. For example, two CCD linear image sensors may inspect the same pattern on one semiconductor wafer, or two CCD linear image sensors inspect two semiconductor wafers provided with the same pattern. May be. Moreover, you may test | inspect with respect to the TFT array substrate used for a liquid crystal display device, a color filter substrate, or a liquid crystal panel. Further, the inspection object may be other than the above.
[0050]
【The invention's effect】
ADVANTAGE OF THE INVENTION According to this invention, the position shift detection method and position shift detection apparatus which can detect the position shift between images correctly in a short time can be provided, and the image processing which correct | amends position shift and performs a comparison process It is preferable to use the method, the image processing apparatus, the inspection method, and the inspection apparatus.
[Brief description of the drawings]
FIG. 1 is a perspective view showing a configuration of an inspection apparatus according to the present invention.
FIG. 2 is a diagram illustrating an example of an image acquired by a CCD linear image sensor.
FIG. 3 is a flowchart showing a processing procedure of an image processing method according to the present invention.
FIG. 4 is a block diagram showing a configuration of an image processing apparatus according to the present invention.
FIG. 5 is a graph showing an example of signal processing of image data.
FIG. 6 is a graph showing a state of signal processing for detecting a displacement amount according to the first embodiment of the present invention;
FIG. 7 is a graph showing a state of signal processing for detecting a displacement amount according to the second embodiment of the present invention.
FIG. 8 is a perspective view showing a configuration of a conventional defect inspection apparatus.
[Explanation of symbols]
1 CCD linear image sensor
2 Substrate
3 Image processing device
4 images
5 Wiring
31 Projection Waveform Generation Circuit
32 FIFO memory
33 DSP
34 Deviation correction circuit
35 Image memory
36 Image comparison circuit

Claims (13)

An inspection method for inspecting an object to be inspected having a plurality of identical design patterns based on a comparison processing result between first image data and second image data obtained by imaging the same design pattern at different positions. ,
(A) a first computing means Fourier transforming the one-dimensional data obtained from the first image data and the second image data ;
(B) Based on the result of the step (a) , the second calculation means respectively phase of the first image data and the second image data at a frequency extracted from an unspecified frequency under a certain condition. A step of seeking
(C) third arithmetic means, said determined at step (b), the phase of the frequency of the first image data, based on the difference between the phase of the frequency of the second image data, wherein Obtaining a displacement amount between the first image data and the second image data ;
(D) a shift correction circuit correcting the positional shift amount obtained in step (b) with respect to the first image data or the second image data;
Inspection method having The first calculation means obtains coefficients corresponding to a plurality of frequencies by the Fourier transform in the step (a),
In the step (b), the second calculation means obtains an intensity at each frequency of the plurality of frequencies, extracts a frequency based on an extreme value of the intensity, and obtains a phase at the extracted frequency. The inspection method according to claim 1. A projection waveform generation circuit acquiring two-dimensional data based on the first image data;
The inspection method according to claim 1, wherein the projection waveform generation circuit includes a step of generating projection waveform data as the one-dimensional data based on the two-dimensional data. The step of generating the projection waveform data by the projection waveform generation circuit generates one-dimensional projection waveform data for a plurality of different directions,
The inspection method according to claim 3, wherein the steps (a), (b), and (c) are performed on each of the one-dimensional projection waveform data. (E) an image comparison circuit comparing the image data corrected in the step (d);
(F) outputting an inspection signal based on the comparison processing result;
The inspection method according to any one of claims 1 to 4, further comprising:   An inspection apparatus that performs an inspection based on a comparison processing result between first image data and second image data obtained by imaging the same design pattern at different positions of a plurality of objects to be inspected having the same design pattern. ,
  An imaging means for capturing an image of the inspection object;
  First computing means for Fourier transforming one-dimensional data obtained from the first image data and the second image data;
  Based on the result of the first calculation, second calculation means for obtaining respective phases of the first image data and the second image data at frequencies extracted from unspecified frequencies under a certain condition;
  Based on the difference between the phase of the first image data at the frequency and the phase of the second image data at the frequency obtained in the second calculation, the first image data and the second image data Calculation means for performing a third calculation to obtain a positional deviation amount between the image data and
  A displacement correction circuit for correcting the displacement amount obtained by the displacement amount detection device with respect to the first image data or the second image data;
  An inspection apparatus comprising:   In the first calculation, a coefficient corresponding to a plurality of frequencies is obtained by the Fourier transform,
  6. The inspection according to claim 5, wherein, in the second calculation, an intensity at each frequency of the plurality of frequencies is obtained, a frequency is extracted based on an extreme value of the intensity, and a phase is obtained at the extracted frequency. apparatus.   The inspection apparatus according to claim 5, further comprising a projection waveform generation circuit that generates projection waveform data as one-dimensional data based on the two-dimensional data, wherein the first image data is two-dimensional data.   The projection waveform generation circuit generates one-dimensional projection waveform data for a plurality of different directions,
  The inspection apparatus according to claim 7, wherein the first calculation, the second calculation, and the third calculation are executed for each of the one-dimensional projection waveform data.   The inspection apparatus according to claim 7 or 8, wherein the projection waveform generation circuit generates a projection waveform by integrating the two-dimensional data in a first direction.   11. The projection waveform generation circuit according to claim 7, wherein differentiation processing for differentiating the two-dimensional data in a second direction orthogonal to the first direction is performed before the projection waveform is generated. The inspection device described.   The inspection apparatus according to claim 5, further comprising a storage unit that stores the displacement amount.   An image comparison circuit for comparing the corrected image data;
  The inspection apparatus according to any one of claims 6 to 12, which outputs an inspection signal based on the comparison result.

Images