Abstract: The present disclosure proposes a method and apparatus for multi-model joint denoising training. Multiple models may be obtained. A set of training samples may be denoised through the multiple models. The multiple models may be trained with the set of denoised training samples.

Abstract: Methods and systems for setting up alignment of a specimen are provided. One system includes computer subsystem(s) configured for acquiring two-dimensional (2D) images generated from output of a detector of an output acquisition subsystem at template locations in corresponding areas of printed instances on a specimen. The computer subsystem(s) determine offsets in x and y directions between the template locations using the 2D images and determine an angle of the specimen with respect to the output acquisition subsystem based on the offsets. If the angle is greater than a predetermined value, the computer subsystem(s) rotate the specimen and repeat the steps described above. If the angle is less than the predetermined value, the computer subsystem(s) store one of the 2D images for alignment of the specimen in a process performed on a specimen. The 2D images may include multi-mode images, which may be fused prior to determining the offsets.

Abstract: Defects of interest and nuisance can be separated into different segments which enables detection of the defects of interest in only one segment. A region of an image can be segmented into a plurality of segments. A range attribute of the segments can be determined. Thresholding can be used to select one of the segments from the range attribute. The segment that is selected can be dilated.

Abstract: A projection is determined in a semiconductor image, which can be an X projection and/or a Y projection. At least one threshold is applied to the projection thereby forming at least one segment within the region. A fine segment can be determined in the region using a distance value from the projection. Defect detection can be performed in one of the fine segments.

Abstract: Methods and systems for setting up alignment of a specimen are provided. One system includes computer subsystem(s) configured for acquiring two-dimensional (2D) images generated from output of a detector of an output acquisition subsystem at template locations in corresponding areas of printed instances on a specimen. The computer subsystem(s) determine offsets in x and y directions between the template locations using the 2D images and determine an angle of the specimen with respect to the output acquisition subsystem based on the offsets. If the angle is greater than a predetermined value, the computer subsystem(s) rotate the specimen and repeat the steps described above. If the angle is less than the predetermined value, the computer subsystem(s) store one of the 2D images for alignment of the specimen in a process performed on a specimen. The 2D images may include multi-mode images, which may be fused prior to determining the offsets.

Abstract: A projection is determined in a semiconductor image, which can be an X projection and/or a Y projection. At least one threshold is applied to the projection thereby forming at least one segment within the region. A fine segment can be determined in the region using a distance value from the projection. Defect detection can be performed in one of the fine segments.

Abstract: Defects of interest and nuisance can be separated into different segments which enables detection of the defects of interest in only one segment. A region of an image can be segmented into a plurality of segments. A range attribute of the segments can be determined. Thresholding can be used to select one of the segments from the range attribute. The segment that is selected can be dilated.

Abstract: A method includes receiving one or more images of three or more die of a wafer, determining a median intensity value of a set of pixel intensity values acquired from a same location on each of the three or more die, determining a difference intensity value for the set of pixel intensity values by comparing the median intensity value of the set of pixel intensity values to each pixel intensity value, grouping the pixel intensity values into an intensity bin based on the median intensity value of the set of pixel intensity values, generating an initial noise boundary based on a selected difference intensity value in the intensity bin, generating a final noise boundary by adjusting the initial noise boundary, generating a detection boundary by applying a threshold to the final noise boundary, and classifying one or more pixel intensity values outside the detection boundary as a defect.

Abstract: Defect detection on transparent or translucent wafers can be performed on a die using references from the same die. A first calculated value based on a kernel size, such as a moving mean, is determined. A first difference is determined by subtracting the first calculated value from a pixel intensity. Candidate pixels with a first difference above a threshold are classified. A second calculated value based on a kernel size, such as a local median, is determined. A second difference is determined by subtracting the second calculated value from the pixel intensity. Pixels that include a defect are classified when the second difference is above the threshold.

Abstract: Two or more color data can be combined to form a new data source to enhance sensitivity to defocus signal. Defocus detection can be performed on the newly formed data source. In a setup step, a training wafer can be used to select the best color combination, and obtain defocus detection threshold. This can include applying a segment mask, calculating mean intensities of the segment, determining a color combination that optimizes defocus sensitivity, and generating a second segment mask based on pixels that are above a threshold to sensitivity. In a detection step, the selected color combination is calculated, and the threshold is applied to obtain defocus detection result.

Abstract: Methods and systems for monitoring a non-defect related characteristic of a patterned wafer are provided. One computer-implemented method includes generating output responsive to light from a patterned wafer using an inspection system. The method also includes determining differences between a value of a non-defect related characteristic of the patterned wafer and a known value of the non-defect related characteristic based on differences between one or more attributes of the output and one or more attributes of other output of the inspection system for a different patterned wafer having the known value of the non-defect related characteristic.

Abstract: Defect detection on transparent or translucent wafers can be performed on a die using references from the same die. A first calculated value based on a kernel size, such as a moving mean, is determined. A first difference is determined by subtracting the first calculated value from a pixel intensity. Candidate pixels with a first difference above a threshold are classified. A second calculated value based on a kernel size, such as a local median, is determined. A second difference is determined by subtracting the second calculated value from the pixel intensity. Pixels that include a defect are classified when the second difference is above the threshold.

Abstract: A system, method, and computer program product are provided for identifying fabricated component defects using a local adaptive threshold. In use, images are received for target and reference components of a fabricated device. Additionally, a difference image is generated from the target and reference component images, and defect candidates for the target component are identified from the difference image. Further, for each of the identified defect candidates at a location in the difference image: a threshold is determined based on a local area surrounding the location of the defect candidate, and a signal at the location of the defect candidate is compared to the threshold to determine whether the defect candidate is a defect.

Abstract: Two or more color data can be combined to form a new data source to enhance sensitivity to defocus signal. Defocus detection can be performed on the newly formed data source. In a setup step, a training wafer can be used to select the best color combination, and obtain defocus detection threshold. This can include applying a segment mask, calculating mean intensities of the segment, determining a color combination that optimizes defocus sensitivity, and generating a second segment mask based on pixels that are above a threshold to sensitivity. In a detection step, the selected color combination is calculated, and the threshold is applied to obtain defocus detection result.

Abstract: A method includes receiving one or more images of three or more die of a wafer, determining a median intensity value of a set of pixel intensity values acquired from a same location on each of the three or more die, determining a difference intensity value for the set of pixel intensity values by comparing the median intensity value of the set of pixel intensity values to each pixel intensity value, grouping the pixel intensity values into an intensity bin based on the median intensity value of the set of pixel intensity values, generating an initial noise boundary based on a selected difference intensity value in the intensity bin, generating a final noise boundary by adjusting the initial noise boundary, generating a detection boundary by applying a threshold to the final noise boundary, and classifying one or more pixel intensity values outside the detection boundary as a defect.

Abstract: A system, method, and computer program product are provided for identifying fabricated component defects using a local adaptive threshold. In use, images are received for target and reference components of a fabricated device. Additionally, a difference image is generated from the target and reference component images, and defect candidates for the target component are identified from the difference image. Further, for each of the identified defect candidates at a location in the difference image: a threshold is determined based on a local area surrounding the location of the defect candidate, and a signal at the location of the defect candidate is compared to the threshold to determine whether the defect candidate is a defect.

Abstract: A method for segmenting digitized images includes providing a digitized image, selecting a point with a median enhancement greater than a predefined threshold, wherein a contrast enhancing agent was applied to the subject matter of said digitized image prior to acquisition of said image, defining a shape matrix for the selected point in said image from moments of the intensities in a window of points about said selected point, calculating eigenvalues of said shape matrix, determining an eccentricity of a structure underlying said point from said eigenvalues, and segmenting said image based on said eccentricity values, wherein the steps of defining a shape matrix, calculating eigenvalues of said shape matrix, and determining the eccentricity of the underlying structure are repeated for all points in said image.

Abstract: A method for segmenting digitized images includes providing a digitized image, selecting a point with a median enhancement greater than a predefined threshold, wherein a contrast enhancing agent was applied to the subject matter of said digitized image prior to acquisition of said image, defining a shape matrix for the selected point in said image from moments of the intensities in a window of points about said selected point, calculating eigenvalues of said shape matrix, determining an eccentricity of a structure underlying said point from said eigenvalues, and segmenting said image based on said eccentricity values, wherein the steps of defining a shape matrix, calculating eigenvalues of said shape matrix, and determining the eccentricity of the underlying structure are repeated for all points in said image.