12546948 (P.T.A.B. Nov. 23, 2015)

Ex Parte McAuley et al

Patent Trial and Appeal BoardNov 23, 2015

12546948 (P.T.A.B. Nov. 23, 2015)

UNITED STATES PATENT AND TRADEMARK OFFICE UNITED STATES DEPARTMENT OF COMMERCE United States Patent and Trademark Office Address: COMMISSIONER FOR PATENTS P.O. Box 1450 Alexandria, Virginia 22313-1450 www.uspto.gov APPLICATION NO. FILING DATE FIRST NAMED INVENTOR ATTORNEY DOCKET NO. CONFIRMATION NO. 12/546,948 08/25/2009 Julian McAuley 20090432USNP-XER2318US01 4335 62095 7590 11/24/2015 FAY SHARPE / XEROX - ROCHESTER 1228 EUCLID AVENUE, 5TH FLOOR THE HALLE BUILDING CLEVELAND, OH 44115 EXAMINER CESE, KENNY A ART UNIT PAPER NUMBER 2668 MAIL DATE DELIVERY MODE 11/24/2015 PAPER Please find below and/or attached an Office communication concerning this application or proceeding. The time period for reply, if any, is set in the attached communication. PTOL-90A (Rev. 04/07) UNITED STATES PATENT AND TRADEMARK OFFICE ____________________ BEFORE THE PATENT TRIAL AND APPEAL BOARD ____________________ Ex parte JULIAN MCAULEY, TEOFILO E. DE CAMPOS, GABRIELA CSURKA, and FLORENT PERRONNIN ____________________ Appeal 2013-010226 Application 12/546,948 Technology Center 2600 ____________________ Before JASON V. MORGAN, JOSEPH P. LENTIVECH, and KARA L. SZPONDOWSKI, Administrative Patent Judges. SZPONDOWSKI, Administrative Patent Judge. DECISION ON APPEAL Appeal 2013-010226 Application 12/546,948 2 Appellants appeal 35 U.S.C. § 134(a) from a final rejection of claims 1–20. We have jurisdiction under 35 U.S.C. § 6(b). We reverse. STATEMENT OF THE CASE Appellants’ invention is directed to the classification of images. (Spec. ¶ 1). Claim 1, reproduced below with the disputed limitations in italics, is illustrative of the claimed subject matter: 1. An image classifier comprising: a digital processor configured to perform operations comprising: recursively partitioning an image into a tree of image regions having the image as a tree root and at least one image patch in each leaf image region of the tree; classifying the image regions using at least one classifier; and outputting classification values for the image regions based on the classifying and on weights assigned to the nodes and edges of the tree. The Examiner’s Rejections Appellants seek our review of the following rejections: Claims 1–19 stand rejected under 35 U.S.C. § 102(b) as being anticipated by Winn et al. (US 2008/0075367 A1, published Mar. 27, 2008). Claim 20 stands rejected under 35 U.S.C. § 103(a) as being unpatentable over the combination of Winn and Lempitsky et al. (US 2010/ 0128984 A1, published May 27, 2010). Appeal 2013-010226 Application 12/546,948 3 ANALYSIS Issue: Did the Examiner err in finding Winn discloses “recursively partitioning an image into a tree of image regions having the image as a tree root and at least one image patch in each leaf image region of the tree,” as recited in independent claim 1? In the Final Office Action, the Examiner finds “Winn discusses partitioning an image into a grid and repeatedly labeling and learning the parts of training images using a recursive decision tree.” (Final Act. 5, citing Winn ¶¶ 26, 71, 72, 75; see also Final Act. 2–3). Appellants contend “[t]he partitioning (and the subsequent deformation) disclosed in Winn is not recursive partitioning.” (App. Br. 9). Rather, Appellants contend Winn discloses a flat partitioning scheme. (App. Br. 13, citing Fig. 7) (emphasis omitted). As a result, Appellants contend Winn does not partition an image into “a tree of image regions having the image as a tree root” because “all image regions are ‘at the same level’, i.e. on a single grid.” (App. Br. 9). Appellants further contend “[t]he fact that Winn uses a decision tree of binary tests as its classifier does not fairly suggest recursively partitioning the image to generate a tree of image regions[.]” (App. Br. 13). The Examiner responds “Winn discloses using a decision tree classification algorithm in order to classify regions of an image.” (Ans. 6, citing Winn Fig. 7, ¶ 34). The Examiner interprets Winn’s “binary tree [] based on image features” to be an image tree as claimed. (Ans. 7, citing Winn ¶¶ 77–78). The Examiner describes Winn’s classifier as “a recursive classifier, which recursively and adaptively designs and partitions the tree according to the detected image features[.]” (Ans. 7). The Examiner also Appeal 2013-010226 Application 12/546,948 4 relies on paragraph 28, Figures 7 and 10a–e of Winn, and describes Figures 10b–e as showing “[t]he tree node (p, q) is the first level indicative of the entire parent image while the image is recursively partitioned into image regions (p−1, p+1), (p, q+1), and (p+1, q+1).” (Ans. 8). In Reply, Appellants contend in Winn, the classifier is constructed as a decision tree, which “has nothing to do with the recursive partitioning of the image.” (Reply Br. 3–4, 6, emphasis added). Appellants argue the classifier does not partition the image in Winn; rather, the partitioning is done by using a grid. (Reply Br. 7). Further, Appellants contend Examiner’s interpretation of Winn’s “binary tree” as an image tree is contrary to the language in claim 1 because the leaves in Winn’s tree are “class labels” rather than image regions. (Reply Br. 5–6). Finally, Appellants contend the Examiner’s interpretation of Figures 10b–e is incorrect. (Reply Br. 10). We are persuaded by Appellants’ contentions. Winn discloses an “image of an object is divided into initial parts using a grid.” (Winn ¶ 26; see also Winn ¶ 28 (“an initial division of an image into parts using a regular grid”); ¶ 33 (“applying a grid or other means of dividing the foreground”)). After the image has been partitioned into parts, the initial parts are deformed as described in Winn ¶ 34. Figure 7 of Winn schematically illustrates partitioning an object into a single layer grid and the deformation process. Further, the portions of Winn cited by the Examiner do not support disclosure of recursive partitioning into an image of tree regions having the image as a tree root and at least one image patch in each leaf image region of the tree. Although Winn discloses a classifier that is a decision tree (Winn ¶ 34), we agree with Appellants that this has nothing to do with partitioning Appeal 2013-010226 Application 12/546,948 5 the image into a tree of image regions having the image as a tree root and at least one image patch in each leaf image region of the tree. We further agree with Appellants that the Examiner’s reading of Figures 10b–e is incorrect and does not depict recursive partitioning of an image into a tree of image regions having the image as a tree root. Therefore, we find Winn discloses partitioning an image into a grid, not recursive partitioning as claimed. Therefore, constrained by the record before us, we do not sustain the Examiner’s rejection with respect to independent claim 1. For the same reason, we do not sustain the rejections of independent claims 9 and 18 and dependent claims 2–8, 10–17, and 19, each of which include the same deficiency discussed above with respect to the rejection of claim 1. The Examiner has not shown that the missing limitation discussed above with respect to claim 18, from which claims 20 depends, would have been obvious over Winn and Lempitsky. Therefore, we are constrained by the record to find the Examiner also erred in rejecting claim 20. DECISION For the above reasons, the Examiner’s rejection of claims 1–20 is reversed. REVERSED dw