Uniloc 2017 LLC. v. Netflix, Inc.

N OTE: This disposition is nonprecedential.
United States Court of Appeals
for the Federal Circuit
______________________
UNILOC 2017 LLC.,
Appellant
v.
NETFLIX, INC.,
Appellee
______________________
2021-2085
______________________
Appeal from the United States Patent and Trademark
Office, Patent Trial and Appeal Board in No. IPR2020-
00044.
______________________
Decided: December 15, 2022
______________________
N ATHAN K. CUMMINGS , Etheridge Law Group, South-
lake, TX, argued for appellant. Also represented by J AMES
ETHERIDGE , RYAN S. L OVELESS .
P ATRICK J OHN M C K EEVER , Perkins Coie LLP, San Di-
ego, CA, argued for appellee. Also represented by
MATTHEW COOK BERNSTEIN; D AN L. BAGATELL , Hanover,
NH; T ARA L AUREN K URTIS , Chicago, IL.
______________________
Before D YK, T ARANTO, and HUGHES , Circuit Judges.
Case: 21-2085 Document: 47 Page: 1 Filed: 12/15/2022

-- 1 of 13 --

UNILOC 2017 LLC. v. NETFLIX , INC. 2
HUGHES , Circuit Judge.
Uniloc 2017 LLC appeals a decision by the Patent Trial
and Appeal Board that claims 1, 3, 5, and 6 of U.S. Patent
No. 6,584,229 are unpatentable. Because we reject Uniloc’s
arguments that the Board improperly construed two claim
limitations, we affirm.
I
A
U.S. Patent No. 6,584,229 relates to a method for en-
coding videos. A video is a sequence of moving images, or
“frames.” ’229 patent at 1:27–31; see also Netflix, Inc. v.
Uniloc 2017 LLC, IPR No. 2020–00044, Ex. 1002, Declara-
tion of Eli Saber, Ph.D. in Support of Petition (Saber Decl.)
¶ 30. Frames are commonly defined by unit of “mac-
roblock.” ’229 patent at 1:27–31; Saber Decl. ¶ 32. A mac-
roblock is a two-dimensional square containing a set
number of brightness, color hue, and color saturation val-
ues. Saber Decl. ¶ 32. A typical macroblock is 16 brightness
values wide by 16 brightness values long, with a corre-
sponding number of color hue and saturation values dis-
persed throughout the square. Id. A “pixel” is the unit used
to describe each brightness value. To put it simply: a mac-
roblock is a 16x16 block of pixels, and a frame is a grid of
macroblocks. Id. at ¶¶ 31–33.
Encoding a video requires translating the pictures in
each frame into a compressed code that can be efficiently
stored or transmitted. ’229 patent at 1:17–37. Before the
priority date of the ’229 patent, conventional techniques
were commonly used to separate each frame into a fore-
ground object region and a background region. Id. at 1:21–
25. It was also well-known that working at the macroblock-
level is more efficient than individually coding each of the
256 pixels that make up the macroblock. Id. at 1:51–63.
The ’229 patent purports to claim a more efficient en-
coding method by only coding at the pixel-level when
Case: 21-2085 Document: 47 Page: 2 Filed: 12/15/2022

-- 2 of 13 --

UNILOC 2017 LLC. v. NETFLIX , INC. 3
necessary and, where possible, reusing code for mac-
roblocks in the background area. Id. at 1:66–2:8. The spec-
ification asserts that dividing a video by pixel using “the
conventional region division technique” is “very compli-
cated,” and it is hard to use conventional processes “in real
time.” Id. at 1:51–54. The patent adds that coding and
transmitting by pixel is inefficient. Id. at 1:54–63. It pur-
ports to solve these problems by dividing the frame into two
macroblock-based regions. Id. at 1:66–2:21. One region is a
stationary background region, which contains the portion
of the image that remains the same from one frame in the
sequence to the next. Id. The other is the moving object re-
gion, which contains the portion of the image that changes
from the previous frame. Id. New pixel-level coding is only
necessary for the moving object region, and so the code for
the background region can be reused at the macroblock-
level to avoid redundancy. Id. Claim 1 is representative:
1. A method, for use in an [sic] macroblock-based
object oriented coding of a [sic] image signal,
wherein the image signal has a stationary back-
ground region and an object region and contains a
current frame and a previous frame, each frame in-
cluding a plurality of macroblocks, comprising the
steps of:
a) dividing the stationary background region and
the object region from an inputted video in a mac-
roblock-by-macroblock basis by using a difference
between the previous frame and the current frame;
b) coding shape information of the object region by
using a known coding technique to generate coded
shape information;
c) coding pixel information of each macroblock con-
tained in the object region by using a selected
known coding technique to generate coded object
pixel information;
Case: 21-2085 Document: 47 Page: 3 Filed: 12/15/2022

-- 3 of 13 --

UNILOC 2017 LLC. v. NETFLIX , INC. 4
d) generating coded pixel information of a previous
frame macroblock corresponding to each current
frame macroblock contained in the stationary back-
ground region as coded stationary pixel infor-
mation; and
e) storing or transmitting coded data coded shape
information, coded object pixel information and
coded stationary pixel information as coded image
signal, and
wherein the step d) includes the step of reusing cor-
responding coded pixel information macroblock
contained in the previous frame without coding the
pixel information of each macroblock contained in
the current frame when a difference between a
pixel value of the macroblock of the current frame
and that of the macroblock of the previous frame in
the same position is identical to or smaller than a
predetermined threshold value.
’229 patent at 3:42–4:21.
B
Appellee, Netflix, Inc., petitioned for inter partes re-
view of independent claim 1 and dependent claims 3, 5, and
6 of the ’229 patent under two obviousness grounds. The
primary art reference in the first ground discloses a
method for segmenting a frame into macroblock-based ob-
ject and background regions. Raj Talluri, et al., A Robust,
Scalable, Object-Based Video Compression Technique for
Very Low Bit-Rate Coding, 7 IEEE T RANSACTIONS ON
CIRCUITS AND SYS . F OR VIDEO T ECH . 1 (Feb. 1997) (Talluri).
Talluri discloses two parts of the segmentation step: (1)
comparing the current frame with the previous frame to
identify the moving objects, and (2) bounding each moving
object region and background region with rectangles and
tiling the rectangles with macroblocks. Talluri at 224. The
resulting regions are macroblock-based after step two, but
Case: 21-2085 Document: 47 Page: 4 Filed: 12/15/2022

-- 4 of 13 --

UNILOC 2017 LLC. v. NETFLIX , INC. 5
Talluri does not require comparing at the macroblock-level
during the first step.
The Board found all challenged claims unpatentable
under the first ground and did not reach the second ground.
J.A. at 37–38. Uniloc appeals, arguing the Board erred in
its claim construction of the “dividing” limitation and the
“selected known coding technique” limitation. We have ju-
risdiction under 28 U.S.C. § 1295(a)(4)(A).
II
Claim construction is a question of law that depends on
underlying findings of fact. Teva Pharms. USA, Inc. v.
Sandoz, Inc., 574 U.S. 318, 325–28 (2015). We review the
Board’s constructions de novo and its factual findings for
substantial evidence. Immunex Corp. v. Sanofi-Aventis US
LLC, 977 F.3d 1212, 1217–18 (Fed. Cir. 2020).
III
A
We begin with the “dividing” limitation of claim 1:
dividing the stationary background region and the
object region from an inputted video in a mac-
roblock-by-macroblock basis by using a difference
between the previous frame and the current frame
. . .
’229 patent at 3:48–51 (emphasis added). The dispute is
whether “macroblock-by-macroblock basis” requires the act
of dividing to be done macroblock-by-macroblock (i.e., one
macroblock at a time); or whether it simply requires that
the dividing results in separate macroblock-based regions.
The Board held the latter: the limitation requires only (1)
using a difference between the previous frame and the cur-
rent frame to (2) result in macroblock-based object and
background regions. J.A. 10–11. We agree.
Case: 21-2085 Document: 47 Page: 5 Filed: 12/15/2022

-- 5 of 13 --

UNILOC 2017 LLC. v. NETFLIX , INC. 6
When construing a claim limitation, we rely first on in-
trinsic evidence and then on extrinsic evidence if neces-
sary. Phillips v. AWH Corp., 415 F.3d 1303, 1317 (Fed. Cir.
2005) (en banc). When the technology is complicated and
the intrinsic record is unclear, expert testimony and other
extrinsic evidence may be decisive. See Wyers v. Master
Lock Co., 616 F.3d 1231, 1240 n.5 (Fed. Cir. 2010) (noting
that expert testimony is “sometimes essential,” especially
“in cases involving complex technology.”). Such is the case
here.
The relevant intrinsic evidence before us is: (1) the
claim language itself, (2) portions of the specification that
describe the dividing step, and (3) portions of the specifica-
tion that explain the purpose of the claimed invention and
how it purportedly solves problems in the prior art.1 After
considering all the evidence de novo, we conclude that the
intrinsic record is ambiguous as to the meaning of “mac-
roblock-by-macroblock basis.”
First, the claim language itself is unclear because it
could be read to support either construction equally. On
one hand, we can imagine how a reader might interpret
1 Netflix asks us to ignore Uniloc’s arguments as re-
lated to the third category of intrinsic evidence, claiming
that such arguments were forfeited. Oral Argument at
12:06. But our forfeiture doctrine does not “preclude a
party from proffering additional or new supporting argu-
ments, based on evidence of record, for its claim construc-
tion.” Interactive Gift Exp., Inv. v. Compuserve Inc., 256
F.3d 1323, 1347 (Fed. Cir. 2001). Here, Uniloc merely cited
additional support in the specification to support the same
argument it had always made—“macroblock-by-mac-
roblock basis,” includes how the dividing occurs, not just
what results from that division. We do not consider Uniloc’s
arguments related to this evidence to be forfeited, and thus
we consider all cited portions of the record.
Case: 21-2085 Document: 47 Page: 6 Filed: 12/15/2022

-- 6 of 13 --

UNILOC 2017 LLC. v. NETFLIX , INC. 7
“macroblock-by-macroblock” to imply comparisons using
“one macroblock-by-one macroblock.” But on the other
hand, the structure of the claim can be read to set forth
only two requirements: a “what” (dividing the stationary
background region and the object region from an inputted
video in a macroblock-by-macroblock basis) and a “how” (by
using a difference between the previous frame and the cur-
rent frame). Given these competing reasonable interpreta-
tions, it is not clear based on the claim language what is
meant by “macroblock-by-macroblock basis.”2
Uniloc disagrees, arguing that the claim language un-
ambiguously requires the dividing step to occur one mac-
roblock at a time. But this asks us to read language into
the claim that is simply not there. The claim language is
not comparing using “one macroblock at a time,” nor is it
comparing “one macroblock-by-one macroblock.” If any-
thing, the fact that the language does not require compar-
ing at the macroblock-level—but instead explicitly requires
comparing between frames—suggests that a person of or-
dinary skill in the art would not read Uniloc’s requirement
into the claim language. ’229 patent at 3:50–51 (requiring
that the division occurs by “using a difference between the
previous frame and the current frame . . .”) (emphasis
added).
Uniloc also argues that the preamble clarifies the
meaning of “macroblock-by-macroblock basis.” The pream-
ble states: “wherein the image signal has a stationary back-
ground region and an object region and contains a current
frame and a previous frame, each frame including a plural-
ity of macroblocks.” ’229 at 3:42–47 (emphasis added).
2 Moreover, while we treat “in a macroblock-by-mac-
roblock basis” as a typographical error, preferring the
phrase “on a macroblock-by-macroblock basis,” this claim
drafting mistake makes already ambiguous claim language
even less clear.
Case: 21-2085 Document: 47 Page: 7 Filed: 12/15/2022

-- 7 of 13 --

UNILOC 2017 LLC. v. NETFLIX , INC. 8
While the preamble may give context to the meaning of a
limitation, it does not make the phrase “macroblock-by-
macroblock basis” any clearer. As the Board acknowledged,
the “previous frame” and “current frame” in the preamble
are the same terms used in the dividing limitation. J.A. 10.
The preamble describes both the current and previous
frames as including “a plurality of macroblocks.” Uniloc ar-
gues this means the first frame must already be defined at
the macroblock level. J.A. 11. But every frame is techni-
cally defined at the macroblock level. The macroblocks ex-
ist as a unit of measuring the frame, no matter if we have
drawn invisible dividing lines around them or not. The
Board’s construction is consistent with the preamble lan-
guage. That the previous frame can be expressed in units
of macroblocks does not require the dividing to be done one
macroblock at a time. Thus, neither the claim language nor
the preamble dictates Uniloc’s narrow interpretation of the
dividing step.
Likewise, the specification also does not clarify what is
meant by “macroblock-by-macroblock basis.” It is worth
noting that nowhere in the specification is this term de-
fined. In fact, the phrase “macroblock-by-macroblock basis”
is only ever used once in the entire patent, and that is in
claim 1.
Moreover, the portions of the specification that de-
scribe the dividing step more generally are still unclear.
The paragraph of the specification cited by Board as de-
scribing the dividing step begins:
The video signal is divided into a macroblock-based
stationary back ground [sic] region and a mac-
roblock-based moving object region by using the
difference between a previous frame of the input
video signal and a current frame of the input video
at step 10.
JA 10–11 (citing ’229 patent at 2:45–49). This language
suggests that the difference between a previous frame and
Case: 21-2085 Document: 47 Page: 8 Filed: 12/15/2022

-- 8 of 13 --

UNILOC 2017 LLC. v. NETFLIX , INC. 9
a current frame need not be determined one macroblock at
a time, so long as the video signal is ultimately “divided
into a macroblock-based stationary back ground [sic] and a
macroblock-based moving object region . . .,” as the specifi-
cation describes.
Uniloc points out that the rest of that paragraph pro-
vides an example by which the dividing step occurs one
macroblock at a time. Opening Br. at 14 (citing ’229 patent
at 2:53–60). But we do not limit the claim language to ex-
amples used in the specification. E.g., Williamson v. Citrix
Online, LLC, 792 F.3d 1339, 1346–47 (Fed. Cir. 2015)
(“This court has repeatedly ‘cautioned against limiting the
claimed invention to preferred embodiments or specific ex-
amples in the specification.’”) (quoting Teleflex, Inc. v. Fi-
cosa N. Am. Corp., 229 F.3d 1313, 1328 (Fed. Cir. 2002)).
The patent describes this example as one of multiple “well-
known techniques,” suggesting there are other methods
that would fall within the dividing limitation. This para-
graph does not dictate dividing one macroblock at a time.
Nor do the portions of the specification about the pur-
pose of the invention and avoiding problems in the prior art
clarify the meaning of “macroblock-by-macroblock basis.”
The specification alludes to at least two purposes of divid-
ing the object region and background region into mac-
roblocks: (1) to make coding more efficient, Responsive Br.
at 16–17 (citing ’229 patent at 1:54–62; 2:4–6; 3:10–19);
and (2) to avoid dividing at the pixel-level, Reply Br. at 4–
5, 16–17 (citing ’229 patent at 1:38–41; 2:2–6). The specifi-
cation then goes on to describe how those purposes are
achieved, starting with the first step: “wherein the in-
putted video is to be divided into an [sic] background region
and a foreground object region by unit of macroblock,” ’229
patent at 2:9–12 (emphasis added). These statements, none
of which use the term “macroblock-by-macroblock basis,”
do not dispositively clarify the meaning of this language.
They can be read to support either interpretation of the
claim language.
Case: 21-2085 Document: 47 Page: 9 Filed: 12/15/2022

-- 9 of 13 --

UNILOC 2017 LLC. v. NETFLIX , INC. 10
Thus, after considering the relevant portions of the in-
trinsic record as a whole, we are left with no clear answer
as to whether a person of ordinary skill in the art would
understand the “dividing” limitation to require dividing
one macroblock at a time. This leaves us to consider the
extrinsic record. Helmsderfer v. Bobrick Washroom Equip.,
Inc., 527 F.3d 1379, 1382 (Fed. Cir. 2008) (affirming the
trial court’s reliance on extrinsic evidence where the intrin-
sic record was not sufficiently clear to determine the plain
meaning of a claim limitation). The only extrinsic evidence
provided by Uniloc is a dictionary definition of “basis,”
which defines the word as “the system or principles accord-
ing to which an activity or process is carried on.” Reply Br.
at 6. We do not find this lone dictionary definition helpful
in clarifying Uniloc’s interpretation.
More helpful is the expert testimony of Dr. Eli Saber,
Netflix’s expert. “Where the Board looks beyond the intrin-
sic evidence and consults extrinsic evidence, we review the
Board’s subsidiary factual findings concerning such extrin-
sic evidence for substantial evidence.” Arista Networks,
Inc. v. Cisco Sys., Inc., 908 F.3d 792, 796 (Fed. Cir. 2018).
The substantial evidence standard is satisfied “if a reason-
able mind might accept the evidence as adequate to sup-
port the finding.” Almirall, LLC v. Amneal Pharms. LLC,
28 F.4th 265, 271–72 (Fed. Cir. 2022).
Here, we hold that substantial evidence supports the
Board’s findings concerning Dr. Saber’s testimony. The
Board credited Dr. Saber’s testimony on claim construc-
tion, citing to his statement about how a person of ordinary
skill (himself) would interpret the term “macroblock-by-
macroblock basis.” J.A. 8 (citing J.A. 1355–56). A reasona-
ble mind could have accepted Dr. Saber’s testimony as ad-
equate to support the finding that a person of skill in the
art would not have read “macroblock-by-macroblock basis”
to require dividing one macroblock at a time. In particular,
Dr. Saber filed two declarations during the proceeding, one
with the petition and one with the reply. Dr. Saber’s first
Case: 21-2085 Document: 47 Page: 10 Filed: 12/15/2022

-- 10 of 13 --

UNILOC 2017 LLC. v. NETFLIX , INC. 11
declaration set forth ample evidence that he qualifies as a
person of ordinary skill in the art and that he reviewed the
claim language and specification through that lens. Saber
Decl. ¶¶ 5–22. His declarations also explained the technol-
ogy and how a person of ordinary skill in the art would un-
derstand that technology, ultimately leading to his claim
language analysis cited by the Board.
Uniloc, in contrast, offered no expert evidence, whether
to explain technological facts or usage in the field that
might help clarify the meaning or implications of the spec-
ification or for any other purpose. In its reply brief on ap-
peal, Uniloc suggests it would defeat the purpose of the
invention to compare the brightness between frames pixel-
by-pixel due to the bandwidth required to do so. But there
is no evidence in the record that a pixel-by-pixel compari-
son of brightness would in fact use excessive bandwidth
and frustrate the goal of the invention. Also important is
that Netflix’s expert’s, Dr. Saber’s, opinions are consistent
with the intrinsic record. As discussed above, the meaning
of “macroblock-by-macroblock basis” is left ambiguous af-
ter a careful review of the intrinsic evidence. This is not a
case where the Board has credited extrinsic evidence that
runs afoul of the claim language or the specification. To the
contrary, the Board cited the specification as additional
support for Dr. Saber’s testimony and the Board’s finding
that “macroblock-by-macroblock basis” did not require di-
viding one macroblock at a time. J.A. 8–9. Thus, the Board
properly relied on Dr. Saber’s testimony to confirm that all
that is required to meet this limitation is (1) using a differ-
ence between the previous frame and the current frame to
(2) result in macroblock-based object and background re-
gions.
We have considered the parties’ remaining arguments
and find them unpersuasive. We affirm as to the “dividing”
limitation.
Case: 21-2085 Document: 47 Page: 11 Filed: 12/15/2022

-- 11 of 13 --

UNILOC 2017 LLC. v. NETFLIX , INC. 12
B
Next, we address the “selected known coding tech-
nique” limitation: “coding pixel information of each mac-
roblock contained in the object region by using a selected
known coding technique to generate coded object pixel in-
formation . . . .” ’229 patent at 4:1–3 (emphasis added).
Uniloc argues that the use of the word “selected” requires
a construction that includes a separate step of selecting
which “known coding technique” to use. Opening Br. at 17–
23. The Board rejected this argument, holding that the
claimed method does not require an additional selecting
step.
We agree with the Board. Similar to our precedent in
Summit 6, LLC v. Samsung Electronics Company, “se-
lected” is “not a step in the claimed method,” but is a
“phrase that characterizes the claimed” use of a known cod-
ing technique to code the pixel information of the object re-
gion. 802 F.3d 1283, 1290–92 (Fed. Cir. 2015). Thus, the
Board did not err by not requiring a separate step of “se-
lecting.”
Uniloc contends that we must give meaning to all the
words in the claim and that using a different term gener-
ally presumes a different meaning. But the fact that the
preceding limitation uses the term “known coding tech-
nique” without using the word “selected” does not import a
new “selecting” requirement into this limitation. It is not
that the word “selected” carries no weight. The known cod-
ing technique must have been selected at some point—just
not as part of the encoding method.
Uniloc also points to step 20 of the specification as sup-
port for the “selecting” requirement. Opening Br. at 19. But
step 20 relates to determining a coding technique for the
shape coding of the object region (limitation 1(b)), not for
pixel coding (limitation 1(c)). The parts of the figure that
appear to relate to limitation 1(c), steps 40 and 50, say
nothing about selecting a particular coding technique.
Case: 21-2085 Document: 47 Page: 12 Filed: 12/15/2022

-- 12 of 13 --

UNILOC 2017 LLC. v. NETFLIX , INC. 13
Steps 40 and 50 only show that one first determines
whether to apply the known coding technique, and then
uses the known coding technique if applicable. There is no
reference in the specification to suggest “selected known
coding technique” requires some unspecified selecting step.
Nor has Appellant cited any prosecution history or extrin-
sic evidence to support this reading of the claim. Appellant
seeks to turn the general presumption that different
phrases have different meanings into a bright line rule that
saves the patent from unpatentability. We have rejected
such an inflexible rule before, and we continue to do so
here. SimpleAir, Inc. v. Sony Ericsson Mobile Commc’ns
AB, 820 F.3d 419, 429 (Fed. Cir. 2016).
We have considered the parties’ remaining arguments
and find them unpersuasive. We affirm as to the “selected
known coding technique” limitation.
AFFIRMED
Case: 21-2085 Document: 47 Page: 13 Filed: 12/15/2022

-- 13 of 13 --