`____________
`
`BEFORE THE PATENT TRIAL AND APPEAL BOARD
`____________
`
`GOOGLE INC.,
`Petitioner,
`
`v.
`
`VEDANTI SYSTEMS LIMITED,
`Patent Owner.
`____________
`
`Case IPR2016-002121
`U.S. Patent No. 7,974,339 B2
`____________
`
`PATENT OWNER’S CONTINGENT MOTION TO AMEND
`PURSUANT TO 35 U.S.C. § 316(d)(1) and 37 C.F.R. § 42.121
`
`
`
`
`
`
`
` Case IPR2016-00215 has been consolidated with this proceeding.
`
` 1
`
`
`
`
`
`Case No. IPR2016-00212
`Patent Owner’s Contingent Motion to Amend
`TABLE OF CONTENTS
`
`SUMMARY ........................................................................................................... 1
`
`ARGUMENT ......................................................................................................... 1
`
`I. The Substitute Claims Do Not Enlarge the Scope of Any Original Claim .... 1
`
`II. The Substitute Claims Do Not Introduce New Matter and Are Supported
`by the Original Disclosures .......................................................................... 2
`
`III. The Substitute Claims Are Patentable Over the Alleged Art ........................ 4
`
`A. Proposed Substitute Claims 14-15 ........................................................... 5
`
`1. The Substitute Claims Are Patentable Over the Art Cited by and to
`the Board ............................................................................................ 5
`
`2. Other References and Alleged Prior Art ........................................... 11
`
`B. Proposed Substitute Claims 16-17 ......................................................... 19
`
`1. The Substitute Claims Are Patentable Over the Art Cited by and to
`the Board .......................................................................................... 19
`
`2. Other References and Alleged Prior Art ........................................... 23
`
`CONCLUSION .................................................................................................... 25
`
`
`
`
`
`i
`
`
`
`Case No. IPR2016-00212
`Patent Owner’s Contingent Motion to Amend
`TABLE OF AUTHORITIES
`
`Statutes
`
`35 U.S.C. § 316 .................................................................................................. 1, 2
`
`Rules
`
`37 C.F.R. § 42.121 ............................................................................................. 1, 2
`
`
`
`
`
`ii
`
`
`
`Case No. IPR2016-00212
`Patent Owner’s Contingent Motion to Amend
`SUMMARY
`
`Patent Owner’s contingent motion to amend seeks the following:
`
`• First, in the event that independent claim 7 is found unpatentable,
`
`substitute claim 14 for claim 7, and substitute dependent claim 15 for
`
`dependent claim 9 (to update the dependency).
`
`• Second, in the event that independent claim 10 is found unpatentable,
`
`substitute claim 16 for claim 10, and substitute dependent claim 17 for
`
`dependent claim 12 (to update the dependency).
`
`The Appendix accompanying this motion sets forth the substitute claims in
`
`markup form to identify the amendments (additions underlined and deletions
`
`stricken). Only one substitute claim is proposed for each of the challenged claims
`
`7, 9, 10, and 12. See 37 C.F.R. § 42.121(a)(3).
`
`ARGUMENT
`
`Patent Owner’s conditional motion presents allowable substitute claims.
`
`I.
`
`The Substitute Claims Do Not Enlarge the Scope of Any Original
`Claim
`
`Each of the substitute claims includes a new limitation in addition to the
`
`limitations found in each respective original claim, satisfying the statutory
`
`requirement that the changes “may not enlarge the scope of the claims of the
`
`patent.” 35 U.S.C. § 316(d)(3); see also 37 C.F.R. § 42.121(a)(2)(ii).
`
`Proposed substitute claim 14 includes all of original independent claim 7’s
`
`
`
`1
`
`
`
`Case No. IPR2016-00212
`Patent Owner’s Contingent Motion to Amend
`limitations as well as the following new limitation: “wherein the optimized matrix
`
`data defines at least two regions having different aspect ratios.” Proposed
`
`substitute dependent claim 15 corresponds to original claim 9. Proposed claim 15
`
`was amended to update the dependency to new claim 14.
`
`Proposed substitute claim 16 includes all of words of original independent
`
`claim 10, including the following underlined additions: dividing an array of pixel
`
`data into two or more regions defined by region data; and selecting a non-
`
`predetermined set of pixel data from each region to produce selection pixel data for
`
`each region. The “non-predetermined” language is a new limitation. The other
`
`added language was necessarily implied by the original text of original claim 10.
`
`Proposed substitute dependent claim 17 corresponds to original claim 12.
`
`Proposed claim 17 was amended to update the dependency to new claim 16.
`
`II. The Substitute Claims Do Not Introduce New Matter and Are
`Supported by the Original Disclosures
`
`The substitute claims do not introduce new matter, 35 U.S.C. § 316(d)(3),
`
`and are supported by the July 16, 2004 original application (Ex. 1002) as well as
`
`the January 16, 2002, PCT application (Ex. 1018), 37 C.F.R. § 42.121(b).
`
`Substitute claim 14 adds a requirement that the optimized matrix data that is
`
`generated “defines at least two regions having non-uniform aspect ratio . . . .” The
`
`original and PCT applications each disclose that, after receiving frame data,
`
`optimized matrix data is generated, and that the matrix data can define regions
`
`
`
`2
`
`
`
`Case No. IPR2016-00212
`Patent Owner’s Contingent Motion to Amend
`having uniform and nonuniform sizes, and symmetrical and nonsymmetrical
`
`shapes. (Exs. 1002 and 1018 at ¶¶ 24, 33, 52-53, 55, 64.) The data is then
`
`transmitted, and the system is able to take the “non-similar matrices” and assemble
`
`them into a frame. (Id. at ¶ 42.) As a result of having nonsymmetrical regions that
`
`vary in size, the regions can have different aspect ratios. For example, as
`
`illustrated in Figure 10 and made explicit in the applications, the regions can be “(7
`
`x 3), (5 x 6), (5 x 4), (7 x 7) , (2 x 3), (2 x 7),” (Id. at ¶ 67 and Fig. 10), which
`
`results in aspect ratios of 7x3, 5x6, 5x4, 1x1, 2x3, and 2x7.
`
`Substitute claim 16 makes three changes to the claims, shown in underlines
`
`below:
`
`dividing an array of pixel data into two or more regions
`defined by region data;
`
`selecting a non-predetermined set of pixel data from each
`region to produce selection pixel data for each region;
`
`The first and third changes do not broaden or narrow the claim: they recite
`
`what was already implied by the claim language. A “region” in original claim 10
`
`is necessarily defined by region data. Indeed, original claim 10 later states that
`
`region data is transmitted. (Ex. 1001 at 12:21.) Also, when “selecting” a “set of
`
`pixel data from each region,” the method of claim 10 necessarily will produce
`
`“selection pixel data for each region.” Indeed, original claim 10 later states that
`
`
`
`3
`
`
`
`Case No. IPR2016-00212
`Patent Owner’s Contingent Motion to Amend
`the selection pixel data is transmitted for each region. (Ex. 1001 at 12:21-22.)
`
`Support for these amendments can be found in various portions of the original
`
`application and the PCT application. (Exs. 1002 and 1018 at ¶¶ 25, 36-40, 46, 47,
`
`57-60, 66, 67.)
`
`The second change adds a requirement that the selection of a set of pixel
`
`data from each region involve “selecting a non-predetermined set of pixel data
`
`from each region . . . .” In addition to the disclosures described above, the
`
`applications disclose selecting pixels from two or more regions. In the
`
`applications’ embodiments, the pixel selection system can select the pixels (1) in
`
`accordance with a “predetermined” sequence or location, (2) randomly, or (3)
`
`using some other suitable selection criteria. (Id. ¶ 25; see also at ¶¶ 37-40, 46, 57-
`
`60.) The amendment excludes the “predetermined” location embodiment—e.g.,
`
`where the system knows beforehand which of the pixels for a given region will be
`
`selected. It encompasses and is at least supported by the random pixel selection
`
`embodiment.
`
`III. The Substitute Claims Are Patentable Over the Alleged Art
`Patent Owner’s Response, submitted today, explains why the original claims
`
`distinguish the prior art relied upon in the IPR. This motion focuses on how the
`
`proposed changes added by the substitute claims impact those arguments.
`
`
`
`4
`
`
`
`Case No. IPR2016-00212
`Patent Owner’s Contingent Motion to Amend
`Proposed Substitute Claims 14-15
`
`A.
`The art cited by and to the Board, the additional art of record, and other art
`
`known to the Patent Owner, alone or in combination, fails to teach or suggest the
`
`new claimed combination. Proposed substitute claim 15 depends from proposed
`
`substitute claim 14 and is patentable over the art cited by and to the Board for at
`
`least the same reasons as proposed substitute claim 14 and further in view of the
`
`additional features recited therein.
`
`1.
`
`The Substitute Claims Are Patentable Over the Art
`Cited by and to the Board
`
`Neither of the primary references relied upon by the Board (Spriggs or
`
`Belfor) include the new limitation added to claim 14: “wherein the optimized
`
`matrix data defines at least two regions having different aspect ratios.” To the
`
`extent that they have “regions,” the regions have uniform aspect ratios.
`
`Spriggs discloses an image transmission system that calculates an estimated
`
`value for all of the picture elements (“pels”) in a block by linearly interpolating the
`
`actual values of the four corner pels of the block. (Ex. 1005 at 2:29–35.) Spriggs
`
`compares the estimated value to the actual values and, when the estimated value
`
`and an actual value differ by more than a threshold value, subdivides the block into
`
`four subblocks. (Id. Figs. 2–3, 6; 2:48–54.) Spriggs repeats this process for each
`
`subblock until subdivision is no longer necessary or possible—or until a minimum
`
`block size is defined. (Id. at 2:51–60.) Spriggs’s blocks and subblocks, however,
`
`
`
`5
`
`
`
`Case No. IPR2016-00212
`Patent Owner’s Contingent Motion to Amend
`have the same aspect ratios. (Id. Figs. 2–3, 6; 3:5–7; 3:21–23; 3:44–46.) Spriggs
`
`does not disclose or even suggest generating optimized matrix data that “defines at
`
`least two regions having different aspect ratios,” as recited in proposed substitute
`
`claim 14.
`
`Belfor discloses a spatially adaptive subsampling method that subdivides an
`
`image into square blocks and, for each square block, assigns a specific
`
`subsampling lattice to represent the pixels to be transmitted for that square block.
`
`(Ex. 1007 at Fig. 4; 1; 4.) However, as Petitioner concedes, Belfor only discloses
`
`uniformly subdividing the image into a plurality of uniformly-sized square blocks
`
`having a uniform aspect ratio—it does not have at least two regions with different
`
`aspect ratios. (Ex. 1029, Grindon Decl. ¶172 (“Belfor teaches only uniform size
`
`matrices.”)) Nowhere does Belfor disclose or even suggest generating optimized
`
`matrix data, or any other data, that “defines at least two regions having different
`
`aspect ratios,” as recited in proposed substitute claim 14.
`
`Patent Owner is, however, aware of prior art disclosing blocks of data
`
`having different aspect ratios. Golin (Ex. 1006) and U.S. Patent No. 4,785,349 to
`
`Keith (Ex. 2015) disclose this feature. Golin is one of the references at issue in
`
`this IPR. Keith was cited by the Examiner in the original prosecution history of
`
`the ’339 patent. Although Golin and Keith are unrelated, they appear to share the
`
`same inventors, and their disclosures for purposes of Patent Owner’s motion are
`
`
`
`6
`
`
`
`Case No. IPR2016-00212
`Patent Owner’s Contingent Motion to Amend
`substantially similar. Golin discloses a digital video transmission system that
`
`detects edges between adjacent pixels whose values differ by more than a threshold
`
`value. (Ex. 1006 at Fig. 18; 19:33–49.) When Golin detects an edge (i.e., when
`
`the values of adjacent pixels in the region differ by more than the threshold value),
`
`Golin splits the region in half, either horizontally or vertically, or in quarters. (Id.
`
`at Figs. 26–27; 13:22–49.) Golin then determines, for each subregion, a fill code
`
`for representing the pixel values of each subregion. (Id. at 13:12–19.) As a result
`
`of splitting regions in half, subregions with different aspect ratios can be obtained.
`
`(Id. at Fig. 27.) For example, subregion 2706 in Figure 27 of Golin and Keith has
`
`a different aspect ratio than the other subregions.
`
`One of ordinary skill in the art would have had no incentive to apply the
`
`non-uniform aspect ratios disclosed by Golin and Keith in Spriggs. The non-
`
`uniform aspect ratios require additional bits to describe the different actions –
`
`horizontal split action, vertical split action and fill action. Ex. 2015, Fig. 30; Ex.
`
`1006, Fig. 30 Whereas Spriggs merely requires a 0 or 1 to describe split or no split.
`
`Thus, when the objective is to reduce data, non-uniform aspect ratio would
`
`increase the necessary division code data. Rather if one were to combine Spriggs
`
`with Golin or Keith, Google suggests in its petition that one of ordinary skill in the
`
`art would substitute the adjacent pixel variation roughness test of Golin for the
`
`interpolation type of roughness test of Spriggs. Paper 2, p. 28. Spriggs’ uniform
`
`
`
`7
`
`
`
`Case No. IPR2016-00212
`Patent Owner’s Contingent Motion to Amend
`aspect ratio subdivision method is retained by the combination proffered by
`
`Google.
`
`Further, as argued in patent owner’s response, there would have been
`
`disincentive to combine the multiple block size approach with the subsampling
`
`lattice mode system of Belfor designed for a uniform block size. Ex. 2001, ¶87, 88
`
`(quoted at p. 21 below) The disincentive to combine Belfor with Golin or Keith
`
`would have been even greater if one were to further complicate the system with
`
`vertical and horizontal rectangles each requiring its own set of modes that would
`
`need to be recognized in decoding.
`
`None of the other art before the Board discloses the new limitation:
`
`“wherein the optimized matrix data defines at least two regions having different
`
`aspect ratios.” Thyagarajan discloses an image compression system for Discrete
`
`Cosine Transform (“DCT”) analysis that divides an image into a plurality of 16x16
`
`blocks and, when the variance of pixel values within a particular 16x16 block is
`
`larger than a threshold, subdivides the block into four uniformly-sized square
`
`subblocks. (Ex. 1008 at Fig. 3A; abstract; 5:35–44; 5:56–57.) More specifically,
`
`Thyagarajan first computes the variance of pixel values within each 16x16 block
`
`and, when the variance of pixel values within a 16x16 block is larger than a
`
`threshold T16, subdivides the 16x16 block into four 8x8 blocks. (Id. at 5:42–50;
`
`5:57–58; 6:9–12.) The same goes for an NxM block. Thyagarajan computes the
`
`
`
`8
`
`
`
`Case No. IPR2016-00212
`Patent Owner’s Contingent Motion to Amend
`variance of pixel values within the block and, when the variance of pixel values
`
`within a particular NxM block is larger than a threshold, the block is subdivided
`
`into four N/2xM/2 blocks.
`
`Thyagarajan’s DCT image compression system merely evaluates blocks
`
`and, in some cases, evenly subdivides the block into four subblocks having the
`
`same uniform aspect ratio as the original. An NxM block splits into four N/2xM/2
`
`blocks. The aspect ratio remains uniform. Nowhere does Thyagarajan disclose or
`
`even suggest generating optimized matrix data, or any other data, that “defines at
`
`least two regions having different aspect ratio,” as recited in proposed substitute
`
`claim 14.
`
`Rostampour discloses image filtering techniques for noise removal which
`
`replace the value of a pixel with the median value of a set of pixels in the pixel’s
`
`“local neighborhood.” (Ex. 1020 at 1–2.) Rostampour’s “local neighborhood” is a
`
`fixed, square window size of 2N+1 by 2N+1 for a predetermined value of N, such
`
`as a 3 by 3 window size for N = 1. (Id. at 2–3.)
`
`Sarver discloses a block adaptive interpolative coding technique for image
`
`compression. (Ex. 1024 at abstract; 7:58–60.) Sarver begins by dividing an image
`
`into uniformly-sized square MxM pixel blocks. (Id. at 7:67–68; 9:3–5.) Sarver
`
`then predicts pixel values for the current MxM block under consideration by
`
`linearly interpolating pixel values from neighboring blocks. (Id. at 7:68–8:46.)
`
`
`
`9
`
`
`
`Case No. IPR2016-00212
`Patent Owner’s Contingent Motion to Amend
`Finally, Sarver computes the residual error for the MxM block and, when the
`
`residual error for the block exceeds a predefined threshold Te, repeatedly
`
`subdivides the block into four uniformly-sized subblocks until the residual error
`
`falls below the threshold or the subblock size reaches the minimum block size of
`
`1x1. (Id. at Fig. 4; 8:47–59; 9:3–10; 12:54–57; 13:21–30.)
`
`Brown discloses a compression scheme for compressing audio and video
`
`data that divides an image into uniformly-sized, square MxM blocks of pixels and
`
`tests each block to determine a compression scheme for that block. (Ex. 1025 at
`
`abstract; 2:53–3:24.)
`
`Le discloses a method for seeding and segmenting an image. (Ex. 1026 at
`
`abstract.) Starting with the original undivided image, Le determines the contrast of
`
`a rectangular block by calculating the difference between the minimum and
`
`maximum luminance values of the pixels in the block and, when the contrast
`
`exceeds a predetermined threshold value and the size of the block exceeds a
`
`predetermined minimum size, repeatedly splits the rectangular block into four
`
`uniformly-sized rectangular subblocks. (Id. at Figs. 4B, 7E; 11:53–12:39; 13:11–
`
`31; 19:43–20:5.)
`
`Held discloses various image compression techniques, such as JPEG and
`
`GIF, which subdivide an image into uniformly-sized square blocks of pixels
`
`having the same or similar pixel characteristics. (Ex. 1027 at Fig. 7.1; 13.)
`
`
`
`10
`
`
`
`Case No. IPR2016-00212
`Patent Owner’s Contingent Motion to Amend
`Shi discloses differential pulse code modulation (“DPCM”) techniques for
`
`image and video coding pertaining to transformation, quantization, and codeword
`
`assignment. (Ex. 1028 at 17.)
`
`Rostampour, Sarver, Brown, Le, Held, and Shi at best disclose, if at all,
`
`subdividing a block into subblocks having a uniform aspect ratio, not two or more
`
`regions with different aspect ratios.
`
`2. Other References and Alleged Prior Art
`Keith (Ex. 2015), described above, is a prior art reference that was not
`
`before the Board that illustrates non-uniform aspect ratios. This section discusses
`
`other references that are not before the Board.
`
`The ’339 patent’s prosecution history includes a number of additional
`
`references. But the Patent Owner does not believe that any disclose “generating
`
`optimized matrix data from the frame data, wherein the optimized matrix data
`
`defines at least two regions having different aspect ratios,” as recited in proposed
`
`substitute claim 14. For example, Kafri (U.S. Patent No. 4,776,013) discloses an
`
`image encryption method that uses a master grid to convert an image to an image
`
`grid composed of an image matrix of pixel intensity values. (Ex. 2013 at Fig. 1;
`
`abstract; 3:7–24.) More specifically, Kafri generates, for each pixel in the image, a
`
`“superpixel” that encodes a distinct number of grey levels using a predetermined
`
`combination of duplicated pixels, complemented pixels, and random pixels. (Id. at
`
`
`
`11
`
`
`
`Case No. IPR2016-00212
`Patent Owner’s Contingent Motion to Amend
`5:17–7:39.) For example, Kafri discloses generating, for each pixel in an image, a
`
`2x2 superpixel that provides nine grey levels. (Id. at Fig. 8; 6:31–33.)
`
`In another example, Barnes (U.S. Patent No. 7,050,639) discloses an image
`
`data compression and decompression technique that divides an image data stream
`
`into subregions of differing lengths. (Ex. 2014 at Fig. 4; abstract; 5:51–53; 6:40–
`
`46; 6:50–54; 7:3–9. Barnes then analyzes each subregion of the image data stream
`
`to identify the relative variation in the pixel intensities, or entropy, within each
`
`subregion and select a predetermined compression algorithm for each subregion as
`
`a function of the identified entropy for that region. (Id. at abstract; 2:22–35; 6:57–
`
`61; 7:3–15.) In other words, Barnes merely subdivides an image data stream into
`
`one-dimensional subregions corresponding to the individual rows of an image.
`
`Barnes then selects a compression code for each subregion and compresses each
`
`subregion according to the selected compression code.
`
`The Patent Owner is also aware of International Telecommunication Union
`
`Recommendation H.262. (Ex. 2012.) The H.262 standard discloses a generic
`
`video coding scheme that divides a frame into 16x16 macroblocks comprised of
`
`four 8x8 blocks of luminance samples and two, four, or eight 8x8 blocks of
`
`chrominance samples. (Id. §§ 3.85, 6.1.3; Figs. 6-10, 6-11, 6-12.) Accordingly,
`
`the standard discloses subdividing a frame into blocks having a uniform aspect
`
`ratio. There is no disclosure in the standard for generating optimized matrix data
`
`
`
`12
`
`
`
`Case No. IPR2016-00212
`Patent Owner’s Contingent Motion to Amend
`that “defines at least two regions having non-uniform aspect ratio,” as recited in
`
`proposed substitute claim 14.
`
`The International Search Report for PCT/US02/00503 (to which the ’339
`
`patent claims priority) is attached as Exhibit 2016. The report includes references
`
`to Accad (U.S. Patent No. 5,982,937), Miller (U.S. Patent No. 6,108,383), de
`
`Queiroz (U.S. Patent No. 6,334,001), and Iverson (U.S. Patent No. 5,459,486).
`
`(Exs. 2016-2020.)
`
`None of these references discloses “generating optimized matrix data from
`
`the frame data, wherein the optimized matrix data defines at least two regions
`
`having different aspect ratios,” of claim 14. Rather, those references that disclose
`
`subdividing a block into subblocks have subblocks with uniform aspect ratios as
`
`discussed below.
`
`• Accad discloses a method and an apparatus for compressing a raster
`
`page that separates (i) text and line art objects into a first raster data
`
`layer and (ii) image and photo objects into a second raster data layer.
`
`(Ex. 2017 at Figs. 4a, 4b; 6:35–7:10; 7:51–8:9.) Accad subjects the
`
`first raster data layer of text and line art objects to a lossless
`
`compression procedure such as run-length encoding. (Id. at Fig. 4a;
`
`10:13–28.) In contrast, Accad subjects the second raster data layer of
`
`image and photo objects to a lossy compression procedure such as
`
`
`
`13
`
`
`
`Case No. IPR2016-00212
`Patent Owner’s Contingent Motion to Amend
`JPEG by schematically partitioning the second raster later into
`
`uniformly-sized, square 8x8 blocks of pixels and applying standard
`
`JPEG operations on a row-by-row and block-by-block basis. (Id. at
`
`Fig. 4b; 10:28–11:8.)
`
`• Miller discloses a system that compresses a source image as a series
`
`of uniformly-sized, square 4x4 blocks of pixels and associates each
`
`4x4 block with one of two supertypes, either Intraframe or Interframe,
`
`and one of seven subtypes, either High Res, Medium Res, or Null Res
`
`for Intraframe supertypes or Still, Motion, or Update for Interframe
`
`supertypes. (Ex. 2018 at 1:35–59; cf. 8:17–25 (Miller determines
`
`Chroma Vectors using 2x2 blocks of pixels).) Miller then transforms
`
`the series of 4x4 blocks into a token sequence and compresses the
`
`token sequence using known lossless methods, including Huffman
`
`coding. (Id. at 7:62–67; 15:57–66.)
`
`• De Queiroz discloses an iterative smoothing technique for
`
`compressing mixed raster planes in a mixed raster content (“MRC”)
`
`image that represents a color or gray scale document. (Ex. 2019 at
`
`abstract; 3:28–43.) First, de Queiroz converts an original document
`
`into a pixel map representation composed of uniformly-sized, square
`
`blocks of pixels, such as 8x8 blocks of pixels. (Id. at 5:8–17; 5:46–
`
`
`
`14
`
`
`
`Case No. IPR2016-00212
`Patent Owner’s Contingent Motion to Amend
`54; 6:46–51; 6:60–65.) Next, de Queiroz decomposes, on a block-by-
`
`block basis, the document’s pixel map representation into a three-
`
`plane representation: a reduced-resolution “upper” plane and a
`
`reduced-resolution “lower” plane for JPEG compression; and a high-
`
`resolution binary selector plane for symbol-based compression. (Id. at
`
`5:17–23; 5:54–6:32.) Finally, de Queiroz uses an iterative smoothing
`
`technique to pre-process the reduced-resolution upper and lower
`
`planes using the information contained in the high-resolution binary
`
`selector plane. (Id. at 6:33–65.)
`
`• Iverson discloses a color-mapped display subsystem that combines
`
`palettes of multiple images into a single shared palette by refining the
`
`colors included in the palettes into a set of 256 colors. (Ex. 2020 at
`
`abstract; 4:32–67.)
`
`A foreign counterpart to the ’339 patent is the subject of a nullity proceeding
`
`in Germany. In that proceeding, Google has cited to Exs. 1005 (Spriggs) and 1024
`
`(Sarver), which already were cited in this IPR, and Ex. 2004, which appears to be a
`
`counterpart to Ex. 1025 in this IPR. In addition, Google cited to Exs. 2005-2011.
`
`With respect to Exs. 2006-2008 (D5-D7 in the German proceeding), Google
`
`has not established that the references were published prior to the January 16,
`
`2002, priority date of the ’339 patent. Thus, they are not prior art. For Exhibit
`
`
`
`15
`
`
`
`Case No. IPR2016-00212
`Patent Owner’s Contingent Motion to Amend
`2006 (D5), Google argues that two links show that the reference was published
`
`prior to the priority date. First, Google points to a website: http://wftp3.itu.int/av-
`
`arch/video-site/0109_San/. While Exhibit 2006 (D5) is accessible on the site (via a
`
`link to” VCEG-N83d1.doc”), there is no evidence that the reference was available
`
`on that website prior to January 16, 2002. Second, Google points to an Internet
`
`archive link:
`
`https://web.archive.org/web/20020112043536/http:/standards.pictel.com/ftp/video -
`
`site/0109_San/. That WayBack Machine link purports to be archived on January
`
`12, 2002. But it does not include any link that allows one to access or download
`
`Exhibit 2006—not all the links lead to active pages or downloadable documents. It
`
`is unclear when the document was published. Exhibit 2007 (D6) appears to be a
`
`thesis, and it is unclear whether and when it was indexed or available to the
`
`relevant public. Similarly, it is unclear when and how Exhibit 2008 (D7) became
`
`available.
`
`With respect to Exs. 2005 and 2009-2011 (D4 and D8-D10 in the German
`
`proceeding), none of these references discloses “generating optimized matrix data
`
`from the frame data, wherein the optimized matrix data defines at least two regions
`
`having different aspect ratios,” as recited in proposed substitute claim 14. Rather,
`
`those references that disclose subdividing a block into subblocks have subblocks
`
`with uniform aspect ratios:
`
`
`
`16
`
`
`
`Case No. IPR2016-00212
`Patent Owner’s Contingent Motion to Amend
`• Lee (U.S. Patent No. 5,576,767) (D4) discloses a video compression
`
`system and method that uniformly divides a frame into 32x32 blocks
`
`of pixel data, wherein each 32x32 block is composed of either one
`
`32x32 block, four 16x16 subblocks, sixteen 8x8 subblocks, sixty-four
`
`4x4 subblocks, or a combination of 16x16, 8x8, and 4x4 subblocks.
`
`(Ex. 2005 at Figs. 4a–4b; abstract; 6:55–67.) Lee then performs
`
`motion estimation by replacing a 32x32 base block of pixel data in the
`
`current frame with a composite 32x32 block of predicted pixel data
`
`from the previous frame. (Id. at abstract; 7:1–13.) Specifically, Lee
`
`generates the composite 32x32 block of predicted pixel data by
`
`selecting, from a window of pixel data in the previous frame, the
`
`32x32 block, four 16x16 subblocks, sixteen 8x8 subblocks, sixty-four
`
`4x4 subblocks, or combination of 16x16, 8x8, and 4x4 subblocks of
`
`pixel data that best matches the 32x32 base block of pixel data in the
`
`current frame. (Id. at Fig. 5; 7:1–13.)
`
`• Sullivan (D8) discloses a video compression method that uniformly
`
`divides a frame into 16x16 blocks, 8x8 blocks, or a combination of
`
`16x16 and 8x8 blocks of luminance samples for motion estimation.
`
`(Ex. 2009 at Fig. 3; 9–10.)
`
`• Riley (D9) discloses a video compression method that uniformly
`
`
`
`17
`
`
`
`Case No. IPR2016-00212
`Patent Owner’s Contingent Motion to Amend
`divides a frame into 16x16 macroblocks composed of four 8x8 blocks
`
`of luminance samples, one 8x8 block of red-difference chrominance
`
`samples, and one 8x8 block of blue-difference chrominance samples.
`
`(Ex. 2010 at Fig. 3.13; 3.) Riley further discloses intracoding and
`
`intercoding using either 16x16 blocks or 8x8 blocks. (Id. at 3, 5.)
`
`• Panusopone (D10) discloses a progressive image transmission
`
`technique that uses increasingly dense sampling lattices for each stage
`
`of a multi-stage image transmission. (Ex. 2011 at Fig. 1; Table 1; 1–
`
`5.) Panusopone begins by successively splitting the image using
`
`quad-tree partitioning to separate background regions from detail
`
`regions. (Id. at 2.) Quad-tree partitioning uses square blocks for
`
`subdivisions. Panusopone then analyzes the local characteristics of
`
`each region and allocates different sampling lattices to each
`
`background or detail region. (Id. at 1–2.) Panusopone uses a diluted
`
`lattice with less samples for homogeneous regions in the background
`
`and a dense lattice with more samples for nonhomogeneous regions in
`
`detail areas. (Id. at Table 1; 1–5.) Panusopone then selects vector
`
`quantization to code samples on lattices. (Id. at 3.) Finally,
`
`Panusopone reconstructs the regions by adding the transmitted values
`
`to the estimated value at the right position and then fills the remaining
`
`
`
`18
`
`
`
`Case No. IPR2016-00212
`Patent Owner’s Contingent Motion to Amend
`positions by means of linear interpolation. (Id. at 2.) Thus, Panuspone
`
`uses standard-sized regions for divisions that have the same aspect
`
`ratio.
`
`B.
`
`Proposed Substitute Claims 16-17
`
`The art cited by and to the Board, the additional art of record, and other art
`
`known to the Patent Owner, alone or in combination, fails to teach or suggest the
`
`new claimed combinations. Proposed substitute claim 17 depends from proposed
`
`substitute claim 16 and is patentable over the art cited by and to the Board for at
`
`least the same reasons as proposed substitute claim 16 and further in view of the
`
`additional features recited therein.
`
`Although substitute claim 16 makes three sets of changes to the claims, the
`
`response below focuses on one of them: “selecting a non-predetermined set of
`
`pixel data from each region . . . .”
`
`1.
`
`The Substitute Claims Are Patentable Over the Art
`Cited by and to the Board
`
`One of the primary references relied upon by the Board (Spriggs) fails to
`
`include the new limitation added to claim 16: “selecting a non-predetermined set
`
`of pixel data from each region . . . .” As explained in the accompanying patent
`
`owner response, Spriggs has not been shown to include a process of selecting
`
`pixels from each region. Spriggs does transmit pixels in its encoding process,
`
`however the pixels are always at predetermined locations. As shown in Fig. 6 of
`
`
`
`19
`
`
`
`Case No. IPR2016-00212
`Patent Owner’s Contingent Motion to Amend
`Spriggs, Spriggs first transmits the four corner pixel values for an image and
`
`whenever the image is to be subdivided, Spriggs transmits the same five center
`
`pixel values for the subregions. These pixels are always predetermined.
`
`Q. Right. The selection of the corners is built into the system. It's
`predetermined and will be used to select the pixels for any given
`undivided block?
`A. For an undivided block, yes.
`Grindon Dep. Ex. 2003, 97:10-14. With respect to any given region that is created,
`
`the transmitted pixels for that region are always the four corner pixels according to
`
`Spriggs. Spriggs does not teach a pixel selection system and offers no suggestion
`
`of “selecting a non-predetermined set of pixel data from each region.”
`
`Patent Owner is, however, aware of prior art disclosing non-predetermined
`
`pixel selection. Belfor (Ex. 1006) discloses this feature. As discussed above,
`
`Belfor subdivides an image into uniformly-sized square blocks and, for each
`
`square block, represents the pixels to be transmitted by assigning