Notice of Pre-AIA or AIA Status
The present application, filed on or after March 16, 2013, is being examined under the first inventor to file provisions of the AIA .
Application has canceled claims 3-4, 9-10, 14-15, and 20. Thus, application has pending claims 1-2, 5-8, 11-13, 16-19 and 21-24.
Response to Arguments
Applicant’s arguments, see Remarks page 7, filed 04/23/2026, with respect to claims 6 and 13 have been fully considered and are persuasive. The claim objections of claims 6 and 13 have been withdrawn.
Applicant’s arguments, see Remarks pages 7-9, filed 04/23/2026, with respect to the rejection(s) of claim(s) 1-2, 5-8, 12-13, and 16-19 under 35 U.S.C. 102 have been considered but are moot because of the new ground of rejection. The reference of Yedid was not used for teaching a Mahalanobis distances nor a mean and covariance matrix. The reference of HAN, which was used for now canceled claims 3-4, is relied upon for teaching a Mahalanobis distance and the pixel-wise distribution information comprising a reference mean matrix and a reference covariance matrix (see HAN middle of page 4/23, wherein Mahalanobis distance is used in a comparison for anomaly detection. Additionally, see HAN top of page 5/23, wherein the spectrum sections, i.e., distribution information, which contains the mean vector of the image matrix and the image covariance matrix). In addition, the new reference of ATSUSHI further discloses the group of item images of items for the mean and covariance matrices (see ATSUSHI ¶67-70, wherein a mean vector, i.e., mean matrix, and a variance-covariance matrix are used to calculate color conversion for each pixel. This is done for a plurality of semiconductor images, i.e., group of item images, as described in ¶74). Therefore, the new grounds rejection under 35 U.S.C. 103 discloses all the amended limitation of independent claims 1, 8, and 19.
Applicant’s arguments with respect to claim 11 have been considered but are moot because the new ground of rejection does not rely on any reference applied in the prior rejection of record for any teaching or matter specifically challenged in the argument. Therefore, the new reference of Nakanishi, which is replacing SANDRA in light of the new rejection of its parent claims 1 and 8, in combination with Yedid, HAN, and ATSUSHI, discloses the limitation of original claim 11.
Claim Rejections - 35 USC § 112
The following is a quotation of 35 U.S.C. 112(b):
(b) CONCLUSION.—The specification shall conclude with one or more claims particularly pointing out and distinctly claiming the subject matter which the inventor or a joint inventor regards as the invention.
The following is a quotation of 35 U.S.C. 112 (pre-AIA ), second paragraph:
The specification shall conclude with one or more claims particularly pointing out and distinctly claiming the subject matter which the applicant regards as his invention.
Claim 24 recites the limitation “selecting the pixel-wise threshold” in page 6. There is insufficient antecedent basis for this limitation in the claim. Neither claims 1 nor 2 recite the limitation of selecting a pixel-wise threshold. For the purposes of examination, the Examiner will interpret this as being dependent off of claim 5.
Claim Rejections - 35 USC § 103
In the event the determination of the status of the application as subject to AIA 35 U.S.C. 102 and 103 (or as subject to pre-AIA 35 U.S.C. 102 and 103) is incorrect, any correction of the statutory basis (i.e., changing from AIA to pre-AIA ) for the rejection will not be considered a new ground of rejection if the prior art relied upon, and the rationale supporting the rejection, would be the same under either status.
The following is a quotation of 35 U.S.C. 103 which forms the basis for all obviousness rejections set forth in this Office action:
A patent for a claimed invention may not be obtained, notwithstanding that the claimed invention is not identically disclosed as set forth in section 102, if the differences between the claimed invention and the prior art are such that the claimed invention as a whole would have been obvious before the effective filing date of the claimed invention to a person having ordinary skill in the art to which the claimed invention pertains. Patentability shall not be negated by the manner in which the invention was made.
Claims 1-2, 5-8, 12-13, and 16-19 are rejected under 35 U.S.C. 103 as being unpatentable over Hoshen Yedid WO-2021191908-A1, hereinafter Yedid, in combination with LEI HAN CN-111818330-A, hereinafter HAN, and HIROIKE ATSUSHI JP-2004185259-A, hereinafter ATSUSHI.
As per claim 1, Yedid discloses a method for unsupervised learning based anomaly detection of manufactured items, the method comprises:obtaining multiple item pixels of an item (see Yedid ¶45-47, ¶50-51, and FIG. 1 step 102-104, wherein pixels in the image are acquired);determining item features of the item, based on the multiple item pixels and by a non-item specific neural network (see Yedid FIG. 1 step 104, wherein features of the image are extracted. See also ¶57, wherein extracting features is done with the neural network), the non-item specific neural network is pre-trained to perform feature extraction of objects, at least some of the objects differ from the item (see Yedid ¶39-43, wherein the neural network is pretrained and performs feature extraction. See further ¶51, wherein the feature extraction can be done on multiple training images different from the target image. See also FIGS. 9A-9C, wherein various images different from the anomaly image is displayed); determining, based on the item features, a pixel score for item pixels of the multiple item pixels (see Yedid ¶45-46, wherein an anomaly score for each of the pixels, i.e., pixel score for item pixels, in the target image is calculated using the extracted feature, i.e., item feature, from each feature layer);for each of the item pixels, calculating a distance between the pixel score and reference pixel-wise distribution information (see Yedid ¶45, ¶55, and ¶65, wherein the distance calculation of the neighboring pixels, or cluster, i.e., pixel-wise distribution information, is calculated between the relevant pixel, i.e., pixel score, and the k nearest means cluster); andfor each of the item pixels, determining whether the item pixel is an anomaly pixel based on a comparison between the pixel score and a pixel-wise threshold (see Yedid ¶52-53, wherein the pixel is considered anomalous if the distance exceeds a threshold).
However, Yedid fails to explicitly disclose where HAN teaches:for each of the item pixels, calculating a Mahalanobis distance between the pixel score and reference pixel-wise distribution information (see HAN middle of page 4/23, wherein Mahalanobis distance is used in a comparison for anomaly detection), wherein the reference pixel-wise distribution information comprises a reference mean matrix and a reference covariance matrix (see HAN top of page 5/23, wherein the spectrum sections, i.e., distribution information, which contains the mean vector of the image matrix and the image covariance matrix).
Therefore, it would have been obvious to one of ordinary skill in the art before the effective filing date to modify Yedid’s method by using HAN’s teaching by including Mahalanobis distance to the distance in order to improve the anomaly detection by taking correlations of the dataset, as well as including a mean and covariance matrix to the reference pixel-wise distribution in order to obtain mathematical representation of the distribution.
However, while Yedid, in combination with HAN, discloses a reference mean matrix and a reference covariance matrix, it fails to explicitly disclose where ATSUSHI teaches:the reference pixel-wise distribution information comprises a reference mean matrix and a reference covariance matrix calculated from processing a group of item images of items (see ATSUSHI ¶67-70, wherein a mean vector, i.e., mean matrix, and a variance-covariance matrix are used to calculate color conversion for each pixel. The abnormal value and feature value that are updated from the calculation are then used for the color distribution histogram, i.e., distribution information, further disclosed in ¶73. This is done for a plurality of semiconductor images, i.e., group of item images, as described in ¶74)
Therefore, it would have been obvious to one of ordinary skill in the art before the effective filing date to modify Yedid’s, in combination with HAN, method by using ATSUSHI’s teaching by including a group of item images of items to the mean matrix and a reference covariance matrix in order to obtain further matrix information by grouping all the item images.
As per claim 2, Yedid, in combination with HAN and ATSUSHI, discloses the method according to claim 1 wherein the obtaining of the multiple item pixels comprises receiving an image and generating a cropped image that comprises the multiple item pixels (see Yedid ¶92, wherein the input images are received and cropped to 224x224. See also prior ¶85, DIOR, wherein the bounding box pixels of the object, or item, are resized, i.e., cropped, wherein the object/item is in the image that comprises pixels of the object/item).
As per claim 5, Yedid, in combination with HAN and ATSUSHI, discloses the method according to claim 1 wherein the pixel-wise threshold is selected out of multiple thresholds by conducting an iterative process and are based on one or more anomaly detection parameters (see Yedid ¶157, wherein the pixel-level anomaly detection is scanned over a range of thresholds and scanned for false positive anomaly detection, i.e., detection parameter).
As per claim 6, Yedid, in combination with HAN and ATSUSHI, discloses the method according to claim 5 wherein the one or more anomaly detection parameters comprise false positives, true positives and false negatives (see Yedid ¶157, wherein false positive rates are disclosed).
As per claim 7, Yedid, in combination with HAN and ATSUSHI, discloses the method according to claim 5 wherein the one or more anomaly detection parameters comprise image level detection parameters and anomaly level detection parameters (see Yedid ¶157, wherein the pixel-level anomaly detection is scanned over a range of thresholds and scanned for false positive anomaly detection, i.e., detection parameter).
As per claim 8, the rationale provided in claim 1 is incorporated herein. In addition, Yedid, in combination with HAN and ATSUSHI, discloses wherein the group covariance matrix is calculated based on pixel-wise features across the group of item images (see ATSUSHI ¶67-70, wherein an average vector, i.e., mean matrix, and a variance-covariance matrix are used to calculate color conversion for each pixel. This is done for a plurality of semiconductor images, i.e., group of item images, as described in ¶74), and wherein the group mean value matrix is calculated based on mean values of pixel-wise features across the group of item images (see ATSUSHI ¶68-71 wherein the average vector of each pixel in the frame is calculated across frame images).
As per claims 12-13 and 16-19, the rationale provided in claims 1-2, and 5-8 are incorporated herein. Additionally, Yedid discloses a non-transitory computer readable medium (see Yedid ¶7, wherein a non-transitory computer readable medium is disclosed) and the non-transitory computer readable medium of claims 12-13 and 16-19 corresponds to the method of claims 1-2 and 5-8.
Claims 11 is rejected under 35 U.S.C. 103 as being unpatentable over Yedid, in combination with HAN and ATSUSHI, in further view of Hitoshi Nakanishi Iterative Image Inpainting, hereinafter Nakanishi.
As per claim 11, Yedid fails to explicitly disclose where Nakanishi teaches:The method according to claim 8 comprising calculating a value of a pixel-wise threshold in an iterative manner (see Nakanishi Section 2.3, wherein iterative thresholding for a pixel-wise structural similarity index is disclosed) that comprises calculating values of one or more anomaly detection parameters under different candidates of the values of pixel-wise thresholds (see Nakanishi Section 2.3, wherein the detection parameter
a
i
(
x
)
is disclosed, which uses the candidates
x
0
and
x
~
i
at each
i
t
h
iteration, i.e., different candidates, compared to the threshold
u
at each mask
M
i
).
Therefore, it would have been obvious to one of ordinary skill in the art before the effective filing date to modify Yedid’s, in combination with HAN and ATSUSHI, method by using Nakanishi’s teaching by including an iterative threshold for different candidates to the thresholding in order to double check for the anomaly by using multiple thresholds as well as checking multiple items in a dataset.
Claims 24 is rejected under 35 U.S.C. 103 as being unpatentable over Yedid, in combination with HAN and ATSUSHI, in further view of A. Tosun and A. Bener, “Reducing false alarms in software defect prediction by decision threshold optimization,” hereinafter Tosun, and Charlene E. Caefer, et al. “Analysis of false alarm distributions in the development and evaluation of hyperspectral point target detection algorithms,” hereinafter Caefer.
As per claim 24, Yedid, in combination with HAN and ATSUSHI, fails to explicitly disclose where Caefer discloses:The method according to claim 2 wherein selecting the pixel-wise threshold comprises: searching over a range of thresholds between a minimum distance score pixel and a maximum distance score pixel to identify a best performing threshold (see Caefer page 7/35 and FIG. 1, wherein a range of x axis values, thresholds, is used to generate a ROC curve, which is a classification performance at different decision thresholds. A ratio
A
t
h
is used for the maximum area of perfect detection under the ROC curve), wherein the best performing threshold is defined by a threshold that returns a combination of a highest image level detection rate (see Caefer page 7/35, wherein the ROC is calculated and normalized in order to achieve the maximum area possible for perfect detection), a lowest image level false alarm rate (see Caefer pages 7-8/35, wherein the perfect detection is computed to be Eq. 4, where the constant false alarm rate is within the probability of 0 to 0.1)
Therefore, it would have been obvious to one of ordinary skill in the art before the effective filing date to modify Yedid’s, in combination with HAN and ATSUSHI, method by using Caefer’s teaching by including a best performing threshold to the pixel-wise threshold in order to further optimize the pixel-wise threshold by observing a range of thresholds that have the best image level detection rate.
However, Yedid, in combination with HAN, ATSUSHI, and Caefer fails to explicitly disclose where Tosun discloses:wherein the best performing threshold is defined by a highest defect level detection rate, and a lowest defect level false alarm rate (see Tosun pages 2-3/4 Section 3.1, wherein a Naïve Bayes classifier is used to optimize the ROC curve, which is a range of thresholds, in order to get the lowest false alarms with the highest true positives of the defects).
Therefore, it would have been obvious to one of ordinary skill in the art before the effective filing date to modify Yedid’s, in combination with HAN, ATSUSHI, and Caefer, method by using Tosun’s teaching by including a best performing threshold to the pixel-wise threshold in order to further optimize the pixel-wise threshold by observing a range of thresholds that have the best defect detection rate, while maintaining a low false alarm rate.
Allowable Subject Matter
Claims 21-23 are objected to as being dependent upon a rejected base claim, but would be allowable if rewritten in independent form including all of the limitations of the base claim and any intervening claims.
Conclusion
THIS ACTION IS MADE FINAL. Applicant is reminded of the extension of time policy as set forth in 37 CFR 1.136(a).
A shortened statutory period for reply to this final action is set to expire THREE MONTHS from the mailing date of this action. In the event a first reply is filed within TWO MONTHS of the mailing date of this final action and the advisory action is not mailed until after the end of the THREE-MONTH shortened statutory period, then the shortened statutory period will expire on the date the advisory action is mailed, and any nonprovisional extension fee (37 CFR 1.17(a)) pursuant to 37 CFR 1.136(a) will be calculated from the mailing date of the advisory action. In no event, however, will the statutory period for reply expire later than SIX MONTHS from the mailing date of this final action.
Any inquiry concerning this communication or earlier communications from the examiner should be directed to Bradley Obas Felix whose telephone number is (703)756-1314. The examiner can normally be reached M-F 8-5 EST.
Examiner interviews are available via telephone, in-person, and video conferencing using a USPTO supplied web-based collaboration tool. To schedule an interview, applicant is encouraged to use the USPTO Automated Interview Request (AIR) at http://www.uspto.gov/interviewpractice.
If attempts to reach the examiner by telephone are unsuccessful, the examiner’s supervisor, Vincent Rudolph can be reached at 5712728243. The fax phone number for the organization where this application or proceeding is assigned is 571-273-8300.
Information regarding the status of published or unpublished applications may be obtained from Patent Center. Unpublished application information in Patent Center is available to registered users. To file and manage patent submissions in Patent Center, visit: https://patentcenter.uspto.gov. Visit https://www.uspto.gov/patents/apply/patent-center for more information about Patent Center and https://www.uspto.gov/patents/docx for information about filing in DOCX format. For additional questions, contact the Electronic Business Center (EBC) at 866-217-9197 (toll-free). If you would like assistance from a USPTO Customer Service Representative, call 800-786-9199 (IN USA OR CANADA) or 571-272-1000.
/BRADLEY O FELIX/Examiner, Art Unit 2671
/VINCENT RUDOLPH/Supervisory Patent Examiner, Art Unit 2671
Prosecution Insights