Ex Parte Oisel et al

Board of Patent Appeals and Interferences

Feb 28, 2011

10528636 (B.P.A.I. Feb. 28, 2011)

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.
10/528,636 03/22/2005 Lionel Oisel PF020128 7137
24498 7590 02/28/2011
Robert D. Shedd, Patent Operations
THOMSON Licensing LLC
P.O. Box 5312
Princeton, NJ 08543-5312
EXAMINER
BLACK, LINH
ART UNIT PAPER NUMBER
2159
MAIL DATE DELIVERY MODE
02/28/2011 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 BOARD OF PATENT APPEALS
AND INTERFERENCES
____________

Ex parte LIONEL OISEL, BERTRAND CHUPEAU, and
FRANCOIS LE CLERC
____________

Appeal 2009-009929
Application 10/528,6361
Technology Center 2100
____________

Before JOSEPH L. DIXON, JOHN A. JEFFERY, and CAROLYN D.
THOMAS, Administrative Patent Judges.

JEFFERY, Administrative Patent Judge.

DECISION ON APPEAL2

1 Appellants list Application No. 10/528,636 on page 1 of the Appeal Brief
but also has Application No. 10/877,581 in the Brief’s header (see Br. 2-17)
and discusses the real party for Application No. 10/428,973 (see Br. 3).
Based on the issues presented, we will assume that this appeal relates to
Application No. 10/528,636 which has corresponding claims.
2 The two-month time period for filing an appeal or commencing a civil
action, as recited in 37 C.F.R. § 1.304, or for filing a request for rehearing,
as recited in 37 C.F.R. § 41.52, begins to run from the “MAIL DATE”
(paper delivery mode) or the “NOTIFICATION DATE” (electronic delivery
mode) shown on the PTOL-90A cover letter attached to this decision.
Appeal 2009-009929
Application 10/528,636

2
Appellants appeal under 35 U.S.C. § 134(a) from the Examiner’s
rejection of claims 1-3 and 6. The Examiner indicates that claims 4 and 5
contain allowable subject matter. Ans. 7. We have jurisdiction under
35 U.S.C. § 6(b). We reverse.

STATEMENT OF THE CASE
Appellants’ invention involves clustering key images using spatial and
temporal attributes. See generally Spec. 1. Claim 1 is reproduced below
with a key limitation emphasized:
1. Method of clustering images of a video sequence consisting of
shots and represented by a graph-like structure, a node of the graph
representing a shot or a class of shots defined by key images and the
nodes being connected by edges, comprising the following iteration:
selecting an edge ak connecting nodes ni and nj[,]
calculating a potential of node nm,
merging of the two nodes ni and nj, as a function of the distances
between the attributes of the key images defining the class of shots of
node ni and those of the key images defining the class of shots of node
nj and as a function of the temporal distance of these key images,
calculating a potential of each edge connecting the merged node to
another node of the graph previously connected to nodes ni or nj, as a
function of the distances between the attributes of the key images
defining the class of shots of the merged node and those of the key
images defining the class of shots of the other node and as a function
of the temporal distance between these key images, the new class of
shots associated with the merged node comprising the key images of
the classes of shots of the merged nodes, and
merging of the two nodes and validation of the new graph if an
energy of this graph, which is the sum of the potentials of the nodes
and of the edges, is less than the energy of the graph before merging.
Appeal 2009-009929
Application 10/528,636

3
The Examiner relies on the following as evidence of unpatentability:
Yeo US 5,821,945 Oct. 13, 1998
Oguz US 7,054,367 B2 May 30, 2006
(filed Dec. 31, 2001)
Geiger US 7,212,201 B1 May 1, 2007
(filed Sept. 20, 2000)

THE REJECTION
The Examiner rejected claims 1-3 and 63 under 35 U.S.C. § 103(a) as
unpatentable over Yeo, Oguz, and Geiger. Ans. 3-7.4

THE CONTENTIONS
Regarding independent claim 1, the Examiner finds that Yeo teaches
merging nodes ni and nj as a function of the distances between the key
images’ attributes defining the class of shots for nodes ni and nj but fails to
discuss additionally merging these nodes as function of the key images’
temporal distance. Ans. 4. The Examiner relies on Oguz to teach this
missing limitation. Ans. 4-5.

3 The Examiner states “claims 1-2 and 3-6” are rejected under § 103.
Ans. 3. However, the Examiner only discusses claims 1-3 and 6 in the body
of the rejection (Ans. 3-7) and also indicates that claims 4 and 5 are
allowable if rewritten in independent form (Ans. 7). We therefore
presume—as do Appellants (Br. 5)—that the Examiner intended to reject
only claims 1-3 and 6 and only address arguments thereto.
4 Throughout this opinion, we refer to (1) the Appeal Brief filed November
17, 2008, and (2) the Examiner’s Answer mailed January 26, 2009.
Appeal 2009-009929
Application 10/528,636

4
Appellants argue that Yeo does not suggest a temporal distance or
merging nodes according to potentials which are function of temporal
distances. See Br. 11. Appellants also contend that Oguz does not discuss a
temporal distance relating to key images (Br. 9), and that the claim includes
merging nodes as a function of this temporal distance (see Br. 10).
The issue before us, then, is as follows:

ISSUE
(1) Under § 103, has the Examiner erred in rejecting claim 1 by
finding that Yeo, Oguz, and Geiger collectively would have taught or
suggested merging node ni and nj as a function of the temporal distance of
key images that define ni and nj’s class shots?

FINDINGS OF FACT (FF)
1. Yeo discloses video shots that exhibit visual, spatial, and
temporal similarities are clustered into scenes, where each scene contains
one or more shots of similar content. Using the clustering results and
temporal information associated with each shot, the system builds a graph
with nodes representing scenes and edges representing the story’s progress
from one scene to the next. Yeo, col. 5, ll. 32-40.
2. In the context of clustering video shots, Yeo measures shot
similarities or a proximity index based on image attributes, such as color,
spatial correlation, and shape. Yeo describes first grouping shots that are
Appeal 2009-009929
Application 10/528,636

5
most similar together, and then grouping other shots by their proximity
values. Yeo, col. 6, ll. 26-31; col. 8, ll. 44-62; col. 9, ll. 4-13.
3. Yeo also discusses temporal variations in video shots when
camera motions are prominent (e.g., first shot zooms in to Mr. A and a
second shot zooms out to Ms. B). In this case, a representative image is not
sufficient to analyze this image set it represents. Yeo explains the system
chooses a greatly reduced representative set of frames to represent a video
shot so as to reduce computing loads. Yeo also states that clustering is not
confined to only one such representative image. Yeo, col. 8, ll. 9-41.
4. Yeo provides two examples of a hierarchical organization of
shots. Example 1 is a tree representation of shots in time and has a temporal
relation defined by an edge. Example 2 is a directed graph representation of
shots that has cluster relations governed by a temporal ordering of shots
within two clusters. Yeo also discusses the edge relationships induced by
temporal precedence at level 0 is preserved as one moves up the hierarchy.
Yeo, column 4, ll. 1-58.
5. Oguz teaches detecting a scene change in a video sequence by
computing a coincidence coefficient that indicates the degree of coincidence
between significant video frame edges in a current frame and a prior frame
to within a distance of a small number of blocks. The number of blocks can
be determined based on the temporal distance between the current and prior
frames and the motion in the scene in that temporal vicinity. See Oguz, col.
6, ll. 7-35; col. 8, ll. 8-32; Fig. 2.
Appeal 2009-009929
Application 10/528,636

6
ANALYSIS
Based on the record before us, we find error in the Examiner’s
obviousness rejection of claim 1 which calls for, in pertinent part, merging
nodes as a function of the temporal distance of key images that define nodes
ni and nj’s class shots. As the Examiner acknowledges (Ans. 13-14), Yeo
discusses using temporal information when representing shots graphically.
Yeo also clusters or merges nodes (e.g., a node representing a shot) based on
temporal similarities. See FF 1. Yeo, however, provides no more details
regarding these “temporal similarities,” but only describes measuring shot
similarities based on image attributes, such as color, spatial correlation, and
shape. See FF 2. Also, while Yeo discusses temporal variations (see FF 3),
this temporal variation relates to selecting a representative frame set for a
shot to reduce computations—not a temporal distance of these representative
frames for clustering purposes. See id.
The Examiner also cites to Yeo’s discussion (see Ans. 13-14) of
hierarchical shot organizations and discusses edge relationships induced by a
temporal precedence. See FF 4. However, this temporal precedence does
not address a temporal distance of key images, let alone how this precedence
is used to cluster or merge nodes as claim 1 requires. Yeo discloses a
hierarchical tree presentation has a temporal relation defined by an edge.
See id. This temporal relation relates to edges and not a node’s key, and thus
does not teach merging nodes based on this temporal relationship. See id.
Yeo also discusses a directed graph representation of shots that include
Appeal 2009-009929
Application 10/528,636

7
cluster (e.g., nodes) relations governed by a temporal ordering of shots
within two clusters. See id. While Yeo temporal shot ordering within a
node (e.g., a cluster) governs node relationships, Yeo does not teach or
suggest this temporal ordering is a temporal distance of key images or that
the nodes (e.g., clusters) merged as a function of the temporal ordering. We
therefore disagree with the Examiner’s assertion that Yeo’s temporal
ordering of shots “seems to [be] equivalent to ‘the ordering of shots’
temporal values/distance.” Ans. 14.
On the other hand, Oguz detects a scene change (e.g., detecting edges
between scenes or nodes) by computing a coincidence coefficient. See FF 1,
5. Specifically, the coincidence coefficient indicates the coincidences
between significant edges in a current frame and in a prior frame to within a
distance of a small number of blocks, and the blocks are determined using a
temporal distance between the current and prior frames. See id. Thus,
because Oguz’s coincidence coefficient considers the blocks’ distance and
the block distance is determined using a temporal distance between frames,
the temporal distance is also a component used to compute a coincidence
coefficient. See id. Additionally, Appellants admit as much. See Br. 10
(second emphasis added) (“the disclosed temporal distance in Oguz . . . is
used to scale motion . . . and to compute a coincidence coefficient”).
Nonetheless, Oguz only teaches using this temporal distance to detect
when a scene changes. See FF 5. Thus, at best, Oguz suggests calculating a
potential of an edge (e.g., where a scene representing a shot class changes)
Appeal 2009-009929
Application 10/528,636

8
as a function of a temporal distance. We fail to see how this teaching further
teaches merging nodes (e.g., shots or scenes) as a function of this temporal
distance as recited in claim 1. Nor has the Examiner provided adequate
evidence that an ordinarily skilled artisan would have recognized that
Oguz’s ability to detect scene changes based, in part, on a temporal distance
also suggests clustering or merging nodes as a function of this key image
distance. See Ans. 4-5. The Examiner states Oguz’s system allows for
“different categorization/clustering method[s].” Ans. 5. Even so, we do not
find that Oguz teaches or suggests clustering or merging nodes based on the
recited temporal distance is known in the art and thus that merging nodes as
a function of a temporal distance of key images that define node ni and nj’s
class shots as required by claim 1. See Ans. 14.
For the foregoing reasons, Appellants have persuaded us of error in
the obviousness rejection of: (1) independent claim 1 and (2) dependent
claims 2, 3, and 6 for similar reasons.

CONCLUSION
The Examiner erred in rejecting claims 1-3 and 6 under § 103.

Appeal 2009-009929
Application 10/528,636

9
ORDER
The Examiner’s decision rejecting claims 1-3 and 6 is reversed.

REVERSED

rwk

Robert D. Shedd, Patent Operations
THOMSON Licensing LLC
P.O. Box 5312
Princeton, NJ 08543-5312