Sie sind auf Seite 1von 14

Trials@uspto.gov Paper No.

7
571-272-7822 Entered: October 2, 2019

UNITED STATES PATENT AND TRADEMARK OFFICE


____________

BEFORE THE PATENT TRIAL AND APPEAL BOARD


____________

UNIFIED PATENTS INC.,


Petitioner,

v.

VELOS MEDIA, LLC,


Patent Owner.
____________

Case IPR2019-00720
Patent 9,538,205 B2
____________

Before KIMBERLY McGRAW, JASON W. MELVIN, and


AARON W. MOORE, Administrative Patent Judges.

MELVIN, Administrative Patent Judge.

DECISION
Denying Institution of Inter Partes Review
35 U.S.C. § 314
Case IPR2019-00720
Patent 9,538,205 B2

I. INTRODUCTION
Petitioner, Unified Patents, Inc., filed a Petition (Paper 2, “Pet.”)
requesting inter partes review of claims 1–5 (“the challenged claims”) of
U.S. Patent No. 9,538,205 B2 (Ex. 1001, “the ’205 patent”). Patent Owner,
Velos Media, LLC, timely filed a Preliminary Response. Paper 6 (“Prelim.
Resp.”). Pursuant to 35 U.S.C. § 314 and 37 C.F.R. § 42.4(a), we have
authority to determine whether to institute review.
An inter partes review may not be instituted unless “the information
presented in the petition . . . and any response . . . shows that there is a
reasonable likelihood that the petitioner would prevail with respect to at
least 1 of the claims challenged in the petition.” 35 U.S.C. § 314(a). For the
reasons set forth below, we conclude that Petitioner has not shown a
reasonable likelihood it would prevail in establishing the unpatentability of
at least one challenged claim. We, therefore, do not institute inter partes
review.

A. THE ’205 PATENT


The ’205 patent relates to video coding techniques. See Ex. 1001,
1:18–27. The techniques include evaluating successive images for
differences, producing a “residual image” or “prediction residual,” and
performing a frequency transform on the residual image. Id. at 1:35–43. The
transform produces transform coefficients, which are then coded. Id. One
coding approach for the transform coefficients is context-based adaptive
binary arithmetic coding (CABAC). Id. 1:44–47. In some coding
approaches, the transform coefficients are each represented by a binary
“significant_coeff_flag” or “transform coefficient presence/absence flags”
indicating whether a coefficient is nonzero. Id. at 1:53–63. Further,
2
Case IPR2019-00720
Patent 9,538,205 B2

subblocks of coefficients may be considered groups, and a flag assigned to


each group, indicating whether the group contains any nonzero coefficients.
Id. at 2:6–16.
CABAC also involves a “context index” that is assigned to each
coefficient to represent a probability model used for coding. Id. at 1:64–2:5.
The ’205 patent describes deriving the context index based on the whether
adjacent subblock groups have nonzero coefficients. In particular, it
describes the condition when two adjacent subblocks do not contain nonzero
coefficients (i.e., the two adjacent subblocks contain only coefficients
without a value), and responding to that condition by setting the context
index to one of three different values. Id. at 4:33–50.

B. CHALLENGED CLAIMS
Challenged claims 1 and 4 are independent. Claim 1 is illustrative and
is reproduced below:
1. An arithmetic decoding device for decoding coded data of a
transform coefficient obtained by performing frequency
transform on a target image for each unit domain, the
device comprising:
a sub-block coefficient presence/absence flag decoding unit
configured to decode a sub-block coefficient
presence/absence flag indicating whether or not one or
more non-zero transform coefficient are included in each
of a plurality of sub-blocks into which the unit domain is
split; and
a context index deriving unit configured to derive a context
index corresponding to a transform coefficient
presence/absence flag indicating whether or not the
process target transform coefficient is 0, wherein
the context index deriving unit is configured to derive the
context index having one of three different values, in a

3
Case IPR2019-00720
Patent 9,538,205 B2

case where the non-zero transform coefficient is not


included in two or more adjacent sub-blocks which are
adjacent to a process target sub-block.

Id. at 107:30–48. Claim 4 recites parallel limitations, as part of an


“arithmetic coding device.” Id. at 108:28–46.

C. PROPOSED GROUNDS OF UNPATENTABILITY


Petitioner asserts the following grounds of unpatentability, each based
on 35 U.S.C. § 103(a):
Claims Challenged References
1–5 Ji1 and Kumakura 2

1, 2, and 4 Kumakura and Kung 3

3 and 5 Kumakura, Kung, and Ji

Pet. 3. Petitioner also relies on the Declaration of Dr. Vijay Madisetti


(Ex. 1002). See generally Pet. 4–95.

II. DISCUSSION

A. CLAIM CONSTRUCTION
Petitioner proposes constructions for “transform coefficient
presence/absence flag” and “a sub-block coefficient presence/absence flag.”

1
U.S. Pat. Pub. No. 2013/0188726 A1 (pub. July 25, 2013) (Ex. 1005).
2
Toru Kumakura & Shigeru Fukushima, “Non-CE3: Simplified context
derivation for significance map (JCTVC-I0296),” Joint Collaborative
Team on Video Coding (JCT-VC) of ITU-T SG 16 WP 3 and ISO/IEC
JTC 1/SC 29/WG 11 (Geneva, April 27 – May 7, 2012) (Ex. 1006).
3
Wei-Ying Kung et al., “Zone Partition for Significance Map Coding of
Large TU (JCTVC-I0372),” Joint Collaborative Team on Video Coding
(JCT-VC) of ITU-T SG 16 WP 3 and ISO/IEC JTC 1/SC 29/WG 11
(Geneva, April 27 – May 7, 2012) (Ex. 1007).
4
Case IPR2019-00720
Patent 9,538,205 B2

Pet. 11–12. Petitioner also proposes constructions for a number of


limitations “to the extent that the Board determines” that each invokes
35 U.S.C. § 112 ¶ 6. Pet. 12–17. “Patent Owner takes no position on
Petitioner’s constructions” and asserts “they are immaterial to any of Patent
Owner’s arguments.” Prelim. Resp. 28.
We determine that no express constructions are required to resolve the
question of institution. See Nidec Motor Corp. v. Zhongshan Broad Ocean
Motor Co., 868 F.3d 1013, 1017 (Fed. Cir. 2017); Vivid Techs., Inc. v. Am.
Sci. & Eng’g, Inc., 200 F.3d 795, 803 (Fed. Cir. 1999).

B. UNPATENTABILITY

1. Ji and Kumakura
Ji relates to encoding and decoding video. Ex. 1005 ¶ 2. It discusses
block-based processes for transforming pixel data to transform coefficients.
Id. ¶¶ 4–7. Ji employs significance maps, which indicate the positions of
nonzero transform coefficients. Id. ¶ 6. Further, Ji includes multilevel
significance maps, where higher-level maps indicate which groups of lower-
level coefficients contain nonzero significant-coefficient flags. Id. ¶ 60.
Ji discloses encoding significance maps using “context-adaptive
binary arithmetic coding (CABAC).” Id. ¶ 56. Determining the context for
the encoding in Ji “depends on determining and summing the values of the
significant-coefficient flags at neighboring locations.” Id. ¶ 57; accord id.
¶ 63. Because Ji discloses determining context with the “cumulative sum of
the values of the five neighboring flags . . . . there may up to six contexts.”
Id. ¶ 64. Ji also discloses that, “[i]n some instances, the number of contexts
may be capped, for example at 4.” Id.

5
Case IPR2019-00720
Patent 9,538,205 B2

Neighboring flags in Ji are used to select a context from a “context


set.” Id. ¶ 65. Ji discloses that a model may define more than one context set,
for example by defining regions of the transform unit where each region
uses a different context set. Id. ¶ 66–69. Ji’s Figure 4 is reproduced below:

Figure 4 illustrates an example 16×16 transform unit 110 including a first


region 112, a second region 114, and the DC location, which “may be
considered a ‘third region’ in which there is a single context in the context
set.” Id. ¶ 69.
Because Ji’s context determination does not depend on the adjacent
group flags, Petitioner relies on teachings from Kumakura. Pet. 33–34.

6
Case IPR2019-00720
Patent 9,538,205 B2

Kumakura discloses deriving the context index from group flags, rather than
from the individual significance flags. Ex. 1006, 1. Because the context
index for each square in the subset depends only on the adjacent group (as
opposed to other coefficients in the same subset), Kumakura’s method
removes dependency and allows the context indices in the subset “to be
derived in parallel.” Id. Kumakura’s Figure 3 is reproduced below:

Figure 3 illustrates the “referred subsets”—the significant-coefficient group


flags of adjacent subsets, used to determine the context for an element of the
“current subset.” Id. at 2. Kumakura discloses that the adjacent subsets
define a “pattern” and that pattern is used to derive the context for each of
the flags within the current subset. Id. at 2–3, Figs. 4–5. Kumakura states
that its “proposed technique can reduce the time complexity and improve the
throughput of the context index derivation process for significance map.” Id.
at 3.
Petitioner reasons that skilled artisans would have understood
Kumakura’s approach using right- and lower-adjacent subblocks as
7
Case IPR2019-00720
Patent 9,538,205 B2

potentially improving Ji’s three-context-region system. Pet. 34. According to


Petitioner, the combination would “employ three context regions when a
block’s neighbors do not have a non-zero transform coefficient, because
Kumakura teaches that inferring groups using a diagonal boundary more
accurately predicts transform coefficients when both adjacent values are
zero.” Id. at 34–35.
Patent Owner argues that the claim language does not read on
Petitioner’s combination because Ji’s three regions provide context sets, not
context indices. Prelim. Resp. 42–46. Thus, according to Patent Owner, Ji
does not teach deriving a “context index having one of three different
values” as claimed. Id. at 43. We agree that Petitioner fails to make an
adequate showing based on Ji’s disclosure of three regions.
Ji discloses first selecting a context set and then selecting the context
index (identifying the context for a particular flag) from within that set.
Ex. 1005 ¶¶ 65–69. And Ji discloses determining the context index for each
position of a subblock based on the values of five neighboring flags, thus
allowing for up to six possible contexts for each flag. Ex. 1005 ¶ 63–65.
While Ji discloses that, “[i]n some instances, the number of contexts may be
capped,” Petitioner does not explain why a skilled artisan would have set the
number of contexts to three, as claimed. Ji’s disclosure of three regions to
select the context set does not equate to a “context index having one of three
different values.”
Notably, Petitioner does not provide any explanation or justification
for modifying Ji’s disclosures to limit the number of contexts for each flag to
three, as claimed. See Pet. 41–46. Rather, Petitioner asserts that skilled
artisans would have been motivated “to employ Ji’s three context regions

8
Case IPR2019-00720
Patent 9,538,205 B2

when [a block’s] neighbors do not have a non-zero transform coefficient.”


Id. at 45. That assertion, however, relates to using three regions, not an
index with one of three values. Because Petitioner does not assert a
combination in which the system derives a context index having one of three
values, Petitioner’s justification for using Ji’s teachings with Kumakura’s
does not result in the claimed invention. We conclude Petitioner’s reasoning
is inadequate in light of the differences between Ji’s context regions and the
claimed context index.
Patent Owner further challenges the reason Petitioner asserts skilled
artisans would have combined teachings from Kumakura and Ji. Prelim.
Resp. 46–47. Patent Owner points out that Kumakura’s asserted benefit
comes from using only adjacent-subblock group flags to select a pattern,
which defines the context for all flags in a target subblock simultaneously.
Id. at 47 (citing Ex. 1006, 2–3). Kumakura supports Patent Owner’s view,
because the way that its techniques “reduce time complexity and improve
the throughput of the context index derivation process” depends on using
only the adjacent group flags. Ex. 1006, 2–3. Ji, on the other hand, teaches
individually determining context for each flag in a subgroup. Ex. 1005
¶¶ 63–65. Thus, it appears that the asserted combination would not maintain
Kumakura’s parallel determination of subblock contexts. To the extent
Petitioner relies on its declarant’s testimony, we determine that testimony is
not consistent with the references’ plain disclosures.
Accordingly, we conclude Petitioner has not shown a reasonable
likelihood it would prevail with respect to unpatentability of claim 1 over Ji
and Kumakura. Because Petitioner relies on the same assertions for claim 4

9
Case IPR2019-00720
Patent 9,538,205 B2

(Pet. 58) and because the other challenged claims depend from claim 1 or 4,
we reach the same conclusion for all challenged claims.

2. Kumakura and Kung


Petitioner also asserts claims 1, 2, and 4 would have been obvious
over a combination of Kumakura and Kung. Pet. 64–84. Kung discloses
partitioning a subblock into zones for coding, where each zone may
influence how the context is selected for elements within the zone. Ex. 1007,
2. A portion of Kung’s Figure 1 is reproduced below:

Portion (b) of Figure 1 depicts Kung’s proposed partitioning of a subblock


into zone 0, zone 1, and zone 2. Id. Kung describes that zone 0 applies to the
upper-left element and has one context; the remainder of the elements fall in
zone 1 or zone 2, depending on a stepwise-diagonal partitioning based on
distance from the upper-left element. Id. Within zone 1, three contexts may
be applied to an element, depending on the presence of “non-zero

10
Case IPR2019-00720
Patent 9,538,205 B2

coefficients in the neighborhood,” and within zone 2, elements use a context


“same as the first context in zone 1.” Id.
As noted above, claim element 1[c] recites: “the context index
deriving unit is configured to derive the context index having one of three
different values, in a case where the non-zero transform coefficient is not
included in two or more adjacent subblocks which are adjacent to a process
target sub-block.” Petitioner contends this limitation is taught by the
combination of Kumakura and Kung. See Pet. 73–76 (citing Ex. 1002
¶¶ 135–138). Specifically, Petitioner asserts Kumakura teaches the claimed
dependence on adjacent subblock group flags as described above. See supra
at 7. Petitioner recognizes Kumakura does not teach assigning a context with
one of three different values and relies on Kung’s three context regions.
Pet. 75 (citing Ex. 1007, 1–2). Petitioner asserts that “Kung’s three context
regions would be an improvement to Kumakura’s system employing the two
context regions.” Id. at 66; see also id. at 76 (same). Petitioner asserts the
improvement would come from Kung’s algorithm, which “results in
negligible compression loss while decreasing encoding and decoding time.”
Id. at 76.
Patent Owner challenges the combination, arguing that Kung’s three
zones do not represent the claimed context indices having one of three
different values, because the zones define the approach for determining each
context index, rather than the index itself. Prelim. Resp. 51–52 (“[E]ach
zone either has only one context, or the contexts are determined based solely
on the number of ‘non-zero coefficients in the neighborhood’ . . . .”). We
agree with Patent Owner—Petitioner does not adequately explain how the

11
Case IPR2019-00720
Patent 9,538,205 B2

combination would result in deriving a context index having one of three


different values.
Kung discloses multiple zones that define the possible contexts for
coefficients within each zone. Ex. 1007, 2. For example, coefficients in
zone 1 are assigned one of three possible contexts. Id. Although Kung
discloses that the context applied to all coefficients in zone 2 is the same as
the first possible context in zone 1 (and thus that there are three possible
context values for zones 1 and 2), Kung does not limit the context in zone 0
to be the same as one of the contexts in zone 1. See Ex. 1007, 2. Thus,
Kung’s algorithm would result in selecting a context with one of four
different values for the elements of a subblock. Petitioner has provided no
persuasive explanation or reason for limiting Kung to contexts selected from
three possible values for its three zones. Therefore, Petitioner has not
sufficiently shown that the combination of Kumakura and Kung teaches “the
context index having one of three different values” as recited in claim 1 and
similarly in claim 4.
Patent Owner also argues Petitioner has not shown a reason to
combine Kumakura and Kung to arrive at the claimed subject matter. Prelim.
Resp., 53–56. As with the combination of Ji and Kumakura, Petitioner seems
to ignore that Kumakura’s asserted benefit comes from the ability to derive
contexts for a subblock in parallel—by using the group flags of adjacent
subblocks without regard to the coefficient flags within the subblock. See
Ex. 1006, 3. We agree with Patent Owner that Petitioner has ignored that
Kumakura and Kung work in fundamentally different ways. Although it
might be possible to combine their teachings in a beneficial manner,
Petitioner has not adequately explained such an approach. As with the

12
Case IPR2019-00720
Patent 9,538,205 B2

combination of Kumakura and Ji, Petitioner’s declarant’s testimony does not


comport with the references’ plain disclosures.
Accordingly, we conclude Petitioner has not shown a reasonable
likelihood it would prevail with respect to unpatentability of claim 1 over
Kumakura and Kung. Because Petitioner relies on the same assertions for
claim 4 (Pet. 83–84) and because the claim 2 depends from claim 1, we
reach the same conclusion for claims 2 and 4.

3. Kumakura, Kung, and Ji


For its assertions against claims 3 and 5 based on Kumakura, Kung,
and Ji, Petitioner relies on the assertions made against independent claims 1
and 4, respectively. Pet. 86, 93. Thus, for the reasons discussed above, we
conclude Petitioner has not shown a reasonable likelihood it would prevail
with respect to unpatentability over Kumakura, Kung, and Ji.

III. CONCLUSION
For the reasons discussed above, we conclude Petitioner has not
shown a reasonable likelihood of prevailing with respect to at least one
claim.

IV. ORDER
Accordingly, it is
ORDERED that, pursuant to 35 U.S.C. § 314(a) no inter partes
review of the ’205 patent is instituted.

13
Case IPR2019-00720
Patent 9,538,205 B2

PETITIONER:
Trenton Ward
Lionel M. Lavenue
C. Brandon Rash
Justin N. Mullen
Finnegan, Henderson, Farabow, Garrett & Dunner, LLP
trenton.ward@finnegan.com
lionel.lavenue@finnegan.com
brandon.rash@finnegan.com
justin.mullen@finnegan.com

Jonathan R. Bowser
Roshan S. Mansinghani
Unified Patents Inc.
jbowser@unifiedpatents.com
roshan@unifiedpatents.com

PATENT OWNER:
Brent N. Bumgardner
Thomas C. Cecil
Barry J. Bumgardner
Matthew C. Juren
NELSON BUMGARDNER ALBRITTON P.C.
bbumgardner@nbclaw.net
tom@nelbum.com
barry@nelbum.com
matthew@nelbum.com

14

Das könnte Ihnen auch gefallen