Ex Parte Mase

Patent Trial and Appeal Board

Oct 31, 2018

14423051 (P.T.A.B. Oct. 31, 2018)

UNITED STA TES p A TENT AND TRADEMARK OFFICE
APPLICATION NO. FILING DATE
14/423,051 02/20/2015
24395 7590 11/02/2018
WILMERHALE/DC
1875 PENNSYLVANIA AVE., NW
WASHINGTON, DC 20006
FIRST NAMED INVENTOR
RyotaMase
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
ATTORNEY DOCKET NO. CONFIRMATION NO.
2209712.00138US 1 5968
EXAMINER
BLOOM, NATHAN J
ART UNIT PAPER NUMBER
2666
NOTIFICATION DATE DELIVERY MODE
11/02/2018 ELECTRONIC
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.
Notice of the Office communication was sent electronically on above-indicated "Notification Date" to the
following e-mail address(es):
whipusptopairs@wilmerhale.com
teresa.maia@wilmerhale.com
PTOL-90A (Rev. 04/07)
UNITED STATES PATENT AND TRADEMARK OFFICE
BEFORE THE PATENT TRIAL AND APPEAL BOARD
Ex parte RYOT A MASE
Appeal 2018-004031 1
Application 14/423,051
Technology Center 2600
Before JEREMY J. CURCURI, HUNG H. BUI, and
MICHAEL J. ENGLE, Administrative Patent Judges.
CURCURI, Administrative Patent Judge.
DECISION ON APPEAL
Appellant appeals under 35 U.S.C. § 134(a) from the Examiner's
rejection of claims 1-10. Final Act. 1. We have jurisdiction under 35 U.S.C.
§ 6(b ).
1 The Appeal Brief states NEC Corporation is the real party in interest. App.
Br. 2
Appeal2018-004031
Application 14/423,051
Claims 1-9 are rejected under 35 U.S.C. § 103 as obvious over Monti
et al. (US 2010/0284608 Al, published Nov. 11, 2010; "Monti"2) and Higa
et al. ("Multiple Object Detection Using Local Descriptor Clustering Based
on Keypoint Locations," FIT 2012 (The 11th Forum on Information
Technology) H-007). 3 Final Act. 3-12.
Claim 10 is rejected under 35 U.S.C. § 103 as obvious over Monti,
Higa, and Richardson (US 2011/0025853 Al, published Feb. 3, 2011). Final
Act. 12-13.
We affirm.
STATEMENT OF THE CASE
Appellant's invention relates to "a program for accurately identifying
an object in an image." Spec. ,r 1. Claim 1 is illustrative and reproduced
below with disputed limitations in italics:
1. An object identification apparatus comprising:
a memory for storing instructions; and
a processor configured to execute the instructions
to:
determine whether or not input image descriptors
of feature points extracted from an input image and
reference image descriptors of feature points extracted
from a reference image correctly correspond to each
other;
2 In the Final Action dated February 3, 2017, the Examiner refers to the
Monti reference as "Giuseppe." However, that is incorrect and the reference
should be addressed as "Monti."
3 In the Final Action, the Examiner relies on an English translation of Higa.
2
Appeal2018-004031
Application 14/423,051
[ 1] extract an input image divided area descriptor
from a divided area obtained by dividing the input
image;
[2a] acquire a conversion descriptor of the divided
area included in an area in the input image
corresponding to a position of an image area obtained by
performing a geometric transformation for correcting a
geometric deviation between the input image and the
reference image on a prescribed area of the reference
image [2b] when a score based on the number of
combinations of input image and reference image
descriptors determined to correspond correctly is equal
to or larger than a prescribed value;
[3] match the conversion descriptor with a
difference descriptor extracted from the prescribed area
of the reference image; and
output a result of the match.
PRINCIPLES OF LAW
We review the appealed rejections for error based upon the issues
identified by Appellant, and in light of the arguments and evidence produced
thereon. Ex parte Frye, 94 USPQ2d 1072, 107 5 (BP AI 2010) (precedential).
ANALYSIS
THE OBVIOUSNESS REJECTION OF CLAIMS 1-9 OVER MONTI AND HIGA
Contentions
The Examiner finds Monti and Higa teach all limitations of claim 1.
Final Act. 3-7. In particular, the Examiner finds Monti teaches [1] "extract
an input image divided area descriptor from a divided area obtained by
dividing the input image" as recited in claim 1. Final Act. 5 ( citing Monti
,r,r 85-89, 92-95). The Examiner finds also Monti teaches
3
Appeal2018-004031
Application 14/423,051
[2a] acquire a conversion descriptor of the divided area
included in an area in the input image corresponding to a position
of an image area obtained by performing a geometric
transformation for correcting a geometric deviation between the
input image and the reference image on a prescribed area of the
reference image
as recited in claim 1. Final Act. 5---6 (citing Monti ,r,r 95-100, 109-113).
The Examiner then finds the combination of Monti and Higa teaches
"acquire a conversion descriptor. . . [2b] when a score based on the number
of combinations of input image and reference image descriptors determined
to correspond correctly is equal to or larger than a prescribed value" as
recited in claim 1. Final Act. 6-7 (citing Higa Section 3, ,r 2). The Examiner
further finds the combination of Monti and Higa teaches [3] "match the
conversion descriptor with a difference descriptor extracted from the
prescribed area of the reference image" as recited in claim 1. Final Act. 7
( citing Monti ,r 28).
Appellant presents the following principal arguments:
1. Monti does not teach [ 1] "extract an input image divided area
descriptor from a divided area obtained by dividing the input image" as
recited in claim 1 because "[t]he Advisory Action, however, fails to show
where [Monti] discloses or suggests that the image is divided as a result of
the localization [for a local descriptor]." App. Br. 5; see also Reply Br. 2.
11. Higa does not teach [2a] or [2b]:
The Examiner acknowledges that [Monti] does not
disclose [2a] acquiring the conversion descriptor of the divided
area [2b] when a score based on the number of combinations of
input image and reference image descriptors determined to
correspond correctly is equal to or larger than a prescribed value.
(Response at 6.) However, the Examiner contends that Higa
does. Applicant respectfully disagrees. For example, Higa
4
Appeal2018-004031
Application 14/423,051
shows, at best, that when the number of inliers is sufficiently
larger than the number of outliers, a common object between
images exists. See Higa at Fig. 3. However, this teaching does
not suggest [2a] acquiring a conversion descriptor of a divided
area [2b] when a score based on the number of combinations of
input image and reference image descriptors determined to
correspond correctly is equal to or larger than a prescribed value.
The Advisory Action fails to address Applicant's argument
related to this feature.
App. Br. 6; see also Reply Br. 2-3.
111. "[Monti] fails to show [3] matching a conversion descriptor
with a difference descriptor extracted from the prescribed area of the
reference image." App. Br. 7; see also Reply Br. 3--4.
Our Review
Our decision turns on the meaning of the phrase "a divided area
obtained by dividing the input image."
Appellant's Specification does not explicitly define the phrase "a
divided area obtained by dividing the input image." Appellant's
Specification, however, describes examples of divided areas. Spec. ,r,r 39-
40; see also Fig. 6 ( example of divided area generation pattern). According
to the Specification, "[t]he descriptor extracting unit 302 extracts a
descriptor from each of the divided area images generated by the divided
area image generating unit 301 and outputs the extracted descriptor as an
input image divided area descriptor." Spec. ,r 40.
Thus, under the broadest reasonable interpretation standard, we
construe the phrase "a divided area obtained by dividing the input image" as
an area less than the entire image area from which a descriptor may be
extracted. See Spec. ,r,r 39-40.
5
Appeal2018-004031
Application 14/423,051
Having construed the relevant claim language, we now tum to the
issues raised by Appellant's arguments.
Regarding Appellant's argument (i), according to the Examiner,
Monti's Scale Invariant Feature Transform (SIFT) feature extraction
method, which is used for feature extraction on the input image, divides an
image to extract SIFT features, and the SIFT descriptors map to the claimed
divided area descriptors. See Final Act. 5 ( citing Monti ,r,r 85-89, 92-95);
see also Ans. 12 ("[ the claim limitations do] not further require a method of
dividing the image into areas, a size of the areas, a shape of the areas, or
number of areas.").
We agree with the Examiner. Monti's SIFT feature extraction method
teaches [ 1] "extract an input image divided area descriptor from a divided
area obtained by dividing the input image" as recited in claim 1 because
Monti's SIFT descriptors are extracted from an area less than the entire
image area (SIFT descriptors are local descriptors). See Monti ,r 93
( obtaining SIFT keypoint-descriptor pairs of the image); see also Monti ,r 6
( describing SIFT descriptors as local descriptors). The broad claim language
does not further restrict the "divided area" recited in claim 1.
Regarding Appellant's argument (ii),
[ t ]he test for obviousness is not whether the features of a
secondary reference may be bodily incorporated into the
structure of the primary reference; nor is it that the claimed
invention must be expressly suggested in any one or all of the
references. Rather, the test is what the combined teachings of the
references would have suggested to those of ordinary skill in the
art.
In re Keller, 642 F.2d 413,425 (CCPA 1981). Appellant's argument (ii) is
directed to Higa, while the contested findings are based on the collective
6
Appeal2018-004031
Application 14/423,051
teachings of Monti and Higa. For this reason and for reasons further
explained immediately below, we do not see any error in the Examiner's
factual findings disputed by Appellant's argument (ii).
Here, Monti teaches
[2a] acquire a conversion descriptor of the divided area
included in an area in the input image corresponding to a position
of an image area obtained by performing a geometric
transformation for correcting a geometric deviation between the
input image and the reference image on a prescribed area of the
reference image
as recited in claim 1 because Monti's SIFT descriptors are extracted from an
area less than the entire image area (divided area), and because Monti
defines an identifying figure of projection in the input image as a Euclidean
transformation of an identifying figure in the reference image (geometric
transformation). See Final Act. 5-6 (citing Monti ,r,r 95-100, 109-113); see
also Monti, Abstract. We also adopt as our own the Examiner's further
explanation in the Examiner's Answer:
The teachings of [Monti] were relied upon to teach the
determination of the extracted regional descriptors that
correspond to a descriptor of a divided area of the image. [Monti]
was also relied upon to teach the determination of a spatial
transformation type conversion from the divided area descriptors
of one image (model image M of a reference object having SIFT
points al, a2, a3, etc .... , item 30, having article Z, para 0082-
0083), to the divided area descriptors of a second image (sample
image II of Figure 1 is input and has SIFT reference points b 1,
b2, b3, etc ... , See Fig 1 and para 0093---0095). [Monti] did not
teach the determination or comparison of a score, which the
teachings of Higa were relied upon to disclose.
Ans. 13.
Further, Monti and Higa, collectively, teach "acquire a conversion
descriptor ... [2b] when a score based on the number of combinations of
7
Appeal2018-004031
Application 14/423,051
input image and reference image descriptors determined to correspond
correctly is equal to or larger than a prescribed value" as recited in claim 1
because Higa discloses
geometric transformation parameters between the images are
estimated according to RANSAC using the positional
relationship of the corresponding points in each cluster and if a
corresponding point matches with the estimated geometric
transformation parameters, it is regarded as an inlier, and
otherwise regarded as an outlier. A projective transformation
matrix is used for a geometric transformation model. If the
number of inliers is sufficiently large relative to the number of
outliers, it is determined that a common object between the
images exists.
Higa Section 3, ,r 2; see also Final Act. 6-7 (citing Higa Section 3, ,r 2) and
Ans. 13-14. Put another way, Riga's "score" is the difference between the
number of inliers and the number of outliers. When the teachings of Higa are
combined with Monti, Monti's acquired conversion descriptor is acquired
using the "score" technique from Higa. The Examiner provides a reason for
combining the references that is facially rational and supported by evidence
drawn from the record. See Final Act. 6-7 ("The motivation for combining
would have been to correct a known deficiency of the SIFT based local
descriptor method of[Monti] using the local descriptor determination and
matching process of Higa et al (see page 1, Introduction of Higa).").
Appellant has not presented any particularized argument as to why this
reasoning is incorrect.
Finally, regarding Appellant's argument (iii), this argument also does
not show any error in the Examiner's findings because Monti, when
modified in view of Higa as discussed above, teaches the "conversion
descriptor" recited in claim 1, and Monti teaches matching descriptors of the
8
Appeal2018-004031
Application 14/423,051
input image (II) with the reference image (M). See Monti, Abstract
("acquiring a position and an orientation of the identifying figure (Z) with
respect to a first keypoint-descriptor pair of the image (M) having a match
with a second keypoint-descriptor pair of the first image (II); defining, in the
first image (II), an identifying figure of projection as a Euclidean
transformation of the identifying figure (Z), with reference to the first and
second pairs").
We, therefore, sustain the Examiner's rejection of claim 1.
We also sustain the Examiner's rejection of claims 2-9, which are not
separately argued with particularity. See App. Br. 7; see also Reply Br. 4.
THE OBVIOUSNESS REJECTION OF CLAIM 10 OVER MONTI, HIGA, AND
RICHARDSON
Appellant does not present separate arguments for claim 10. See App.
Br. 7; see also Reply Br. 4.
We, therefore, sustain the Examiner's rejection of claim 10 for the
same reasons discussed above when addressing claim 1.
ORDER
The Examiner's decision rejecting claims 1-10 is affirmed.
No time period for taking any subsequent action in connection with
this appeal maybe extended under 37 C.F.R. § 1.136(a)(l).
AFFIRMED
9