`March 8, 2017
`
`trials@uspto.gov
`571-272-7822
`
`
`
`RECORD OF ORAL HEARING
`UNITED STATES PATENT AND TRADEMARK OFFICE
`- - - - - -
`BEFORE THE PATENT TRIAL AND APPEAL BOARD
`- - - - - -
`GOOGLE INC.,
`Petitioner,
`v.
`VEDANTI LICENSING LIMITED,
`Patent Owner.
`- - - - - - -
`Case IPR2016-002121
`Patent 7,974,339 B2
`Technology Center 2400
`Oral Hearing Held: Tuesday, February 14, 2017
`
`Before: MICHAEL R. ZECHER, JUSTIN T. ARBES, and
`JOHN A. HUDALLA, Administrative Patent Judges.
`
`The above-entitled matter came on for hearing on Tuesday,
`February 14, 2017, at 10:00 a.m., in Hearing Room A, taken at the U.S.
`Patent and Trademark Office, 600 Dulany Street, Alexandria, Virginia.
`
`
`1 Case IPR2016-00215 has been consolidated with this proceeding.
`
`
`
`APPEARANCES:
`
`ON BEHALF OF THE PETITIONER:
`
`
`
`
`
`
`
`
`
`
`
`MICHAEL V. MESSINGER, ESQ.
`BRIAN W. LEE, ESQ.
`Sterne, Kessler, Goldstein & Fox PLLC
`1100 New York Avenue, N.W.
`Suite 600
`Washington, D.C. 20005
`202-371-2600
`
`
`
`
`
`
`
`
`ON BEHALF OF THE PATENT OWNER:
`
`
`
`
`
`
`
`
`ROBERT M. ASHER, ESQ.
`BRANDON SCRUGGS, ESQ.
`Sunstein Kann Murphy & Timbers LLP
`125 Summer Street
`Boston, Massachusetts 02110
`617-443-9292
`
`2
`
`
`
`
`
`
`
`
`
`
`
`
`
`P R O C E E D I N G S
`
`Case IPR2016-00212
`Patent 7,974,339 B2
`
`
`
` (10:00 a.m.)
`JUDGE HUDALLA: Please be seated, everyone.
`Good morning. We will hear argument now in IPR2016- 00212,
`pertaining to U.S. Patent Number 7,974,339. Case
`IPR2016-00215 has been consolidated with this proceeding and
`it pertains to the same patent.
`Who do we have appearing for Petitioner today?
`MR. MESSINGER: Michael Messinger, Your
`
`Honor.
`
`Messinger.
`
`JUDGE HUDALLA: Good morning, Mr.
`
`Patent Owner?
`MR. ASHER: Robert Asher.
`JUDGE HUDALLA: Good morning, Mr. Asher.
`Okay. Well, thank you and welcome. And per our
`order dated January 17, 2017, each party will have 50 minutes
`to argue today. And because we have a Motion to Amend
`pending, we're going to have potentially a four-part agenda
`today.
`
`First, we will hear from Petitioner presenting its
`case- in-chief as to the challenged claims. Petitioner can
`
`1
`2
`3
`4
`5
`6
`7
`8
`9
`10
`11
`12
`13
`14
`15
`16
`17
`18
`19
`20
`21
`22
`
`
`
`3
`
`
`
`Case IPR2016-00212
`Patent 7,974,339 B2
`
`
`
`reserve rebuttal time. Patent Owner will then respond to
`Petitioner's case and will present its case regarding the Motion
`to Amend, and Patent Owner may reserve rebuttal time as to
`Petitioner's arguments regarding the Motion to Amend.
`Third, we will have Petitioner using the rest of its
`time to respond to Patent Owner's presentation and, fourth,
`Patent Owner may then use the remaining time to respond to
`Petitioner's arguments regarding the Motion to Amend, if any
`are made.
`
`I will remind the parties that Petitioner bears the
`burden of proving any proposition of unpatentability by a
`preponderance of the evidence. And I will also remind the
`parties that we have a court reporter today transcribing the
`proceedings and the record will become public after the
`hearing.
`
`I also remind the parties to please mention the
`slide number if you are going to be discussing it so that we
`will have some clarity in the record as we go forward.
`So with that I guess, Mr. Messinger, if you would
`like to proceed. Would you like to reserve some time today?
`MR. MESSINGER: Yes. 20 minutes, Your Honor.
`JUDGE HUDALLA: Okay.
`
`
`
`4
`
`1
`2
`3
`4
`5
`6
`7
`8
`9
`10
`11
`12
`13
`14
`15
`16
`17
`18
`19
`20
`21
`22
`
`
`
`Case IPR2016-00212
`Patent 7,974,339 B2
`
`
`
`MR. MESSINGER: May it please the Board.
`Happy Valentine's Day. Let me first give an overview and then
`I will address the two grounds.
`The '339 patent boils down to two steps, dividing a
`frame of image data into different regions based on level of
`detail and then selecting a pixel from each region for
`transmission. This simple sampling was well known in the
`prior art and set out in Google's petition. Go to slide 2.
`The Board correctly instituted trial on two grounds
`of obviousness for the '339 patent, namely that the challenged
`claims are obvious over Spriggs in view of Golin and the
`second ground Belfor in view of Thyagarajan and further in
`view of Golin. None of the argument or evidence Patent
`Owner has presented since disturbs this result.
`Slide 3. For the Spriggs ground Patent Owner
`simply misreads the teachings of Spriggs as laid out in the
`petition. Spriggs plainly teaches both the analysis and the
`pixel selection as claimed, and Spriggs also teaches the pixel
`data, that is the pixel values, and region data, which are the
`division codes and coordinate addresses for the corners that are
`forming a block in Spriggs.
`
`1
`2
`3
`4
`5
`6
`7
`8
`9
`10
`11
`12
`13
`14
`15
`16
`17
`18
`19
`20
`21
`
`
`
`5
`
`
`
`Case IPR2016-00212
`Patent 7,974,339 B2
`
`
`
`And contrary to what the Patent Owner argues, this
`is all laid out in our petition as I will demonstrate when we
`walk through and supported by the testimony, the original
`testimony of Dr. Grindon. For the Belfor ground Patent Owner
`mainly tries to show that there is no motivation to combine the
`references.
`However, as Google showed in its petition, Belfor's
`approach of dividing a frame into uniform size blocks can be
`substituted with Thyagarajan's subdivision of blocks involving
`non-uniform sizes. Belfor even suggests non-uniform size
`blocks, and the Thyagarajan teaching on subdivision, it is
`drawn from the block size assignment as we will get into, and
`that occurs before DCT, Discrete Cosine Transform,
`processing, and what the petition is relying on is that
`subdivision from the block size assignment.
`And then just one point, the combination of
`references applied in the petition also teaches the region data,
`the data receiving system as in claim 1 and the assembling as
`in the method steps of 7 and 10.
`So let me touch on the '339 patent and then I will
`go into the grounds in more detail. So if we look at slide 4,
`fig. 1 shows the data transmission system in one embodiment
`
`1
`2
`3
`4
`5
`6
`7
`8
`9
`10
`11
`12
`13
`14
`15
`16
`17
`18
`19
`20
`21
`22
`
`
`
`6
`
`
`
`Case IPR2016-00212
`Patent 7,974,339 B2
`
`
`
`and includes the frame analysis system 106 and the pixel
`selection system 108.
`There is also the data receiving system 104 which
`has the pixel data system and the display generation system.
`And this sort of selection of the pixels is what allows the
`reduced amount of image data to be transmitted without having
`to be compressed.
`If you look at slide 5, the way the '339 patent
`works is it does divide up a region and there are two
`embodiments. One is in fig. 9 where it is sort of uniform
`regions or uniform matrices in particular in this example in fig.
`9. And in fig. 10 there is an embodiment where the matrices
`can have non-uniform sizes.
`So let's just touch on the claims and then get into
`the grounds and show why they are obvious. Claim 1 on slide
`6, it is a system claim, and this is important, Your Honor, in
`that the language that is used to describe the analysis system
`and the language that is used to describe the pixel selection, it
`is cast in fairly broad language that just recites receiving data
`and then generating data. It is sort of the form of these claims.
`So for the analysis system it is receiving frame
`data and there is generating region data. And with the pixel
`
`1
`2
`3
`4
`5
`6
`7
`8
`9
`10
`11
`12
`13
`14
`15
`16
`17
`18
`19
`20
`21
`22
`
`
`
`7
`
`
`
`Case IPR2016-00212
`Patent 7,974,339 B2
`
`
`
`selection system it receives region data and generates a set of
`pixel data forming a new set of data for transmission. A pretty
`high level. As we will show, we will demonstrate, the prior art
`teaches exactly that.
`So why don't we look at the first ground, slide 9,
`Spriggs in view of Golin. And actually we can jump to slide
`10 in the interest of time. Spriggs discloses a coding process
`for a transmitter that reduces the image data transmitted, and
`you can see that on fig. 5. That's what we relied on in our
`petition.
`
`There is a dispute with respect to the coding
`process and how that is -- the different aspects of that that read
`on the analysis and the pixel selection, but it is very clear and
`both parties agree that is operating in the processor here.
`JUDGE HUDALLA: Can I interrupt you there?
`We are looking at figure 5. Which portions of figure 5 do you
`contend are the analysis system and which portions do you
`contend are the pixel selection system?
`MR. MESSINGER: Yes, with respect to fig. 5,
`both of those aspects of the coding process are carried out by
`the processor 14. You can see the frame store, which is part of
`the analysis and that receives the frame data, but the aspect of
`
`
`
`8
`
`1
`2
`3
`4
`5
`6
`7
`8
`9
`10
`11
`12
`13
`14
`15
`16
`17
`18
`19
`20
`21
`22
`
`
`
`Case IPR2016-00212
`Patent 7,974,339 B2
`
`
`
`the coding process is carried out in the processor 14 and the
`aspect of the pixel selection is carried out by the processor 14.
`As we get into it, there is additional description in
`Spriggs of this sort of process that it goes through that actually
`identifies different aspects of that process correspond to those
`elements. I can step through that. We can see that in the next
`slide, slide 11.
`JUDGE ZECHER: Counselor, can we take you
`back real quick and look at the language of claim 1. It talks
`about analysis on a pixel selection system.
`In light of the specification do you think that those
`two systems are mutually exclusive or can they be one and the
`same?
`
`MR. MESSINGER: I'm sorry, in light of the
`specification?
`JUDGE ZECHER: In light of the specification of
`the '339 patent, do you think that the analysis system and the
`pixel selection system are mutually exclusive systems or can
`they be one and the same?
`MR. MESSINGER: They can -- there is discussion
`in the '339 patent of these systems, like the analysis and the
`
`
`
`9
`
`1
`2
`3
`4
`5
`6
`7
`8
`9
`10
`11
`12
`13
`14
`15
`16
`17
`18
`19
`20
`21
`
`
`
`Case IPR2016-00212
`Patent 7,974,339 B2
`
`
`
`pixel selection, basically operating on a computer. There is
`actually a number of different embodiments the '339 describes.
`If you look, for example, at column 3 of the '339
`patent, lines 3 to 12, it even describes that the software system
`can be one or more objects. It even says lines of code. And so
`our position is that each of these systems can be just different
`lines of code, different functions that are carried out by the
`coding process.
`And the overall '339, there is nothing in it that
`would preclude a processor such as processor 14 from reading
`on the '339 implementation.
`JUDGE ZECHER: Thank you.
`JUDGE ARBES: But, counsel, you would agree
`that you have to be able to distinguish one from the other, even
`if they are different operations performed by the same
`executing program, I have to be able to distinguish A from B,
`right?
`
`MR. MESSINGER: Yes, again, referring to that
`general claim language, it has to read on that claim language
`and that is pretty broad claim language. We did recognize in
`our petition and the evidence that there is two elements there.
`
`
`
`10
`
`1
`2
`3
`4
`5
`6
`7
`8
`9
`10
`11
`12
`13
`14
`15
`16
`17
`18
`19
`20
`21
`
`
`
`Case IPR2016-00212
`Patent 7,974,339 B2
`
`
`
`There is an analysis system and there is a pixel selection
`system.
`
`Patent Owners then argue that somehow there is
`sort of a separately identifiable kind of additional limitation
`that needs to be read into that. We think that is unwarranted
`given the claim language.
`However, even then our petition and the evidence
`of Spriggs shows that the part of the algorithm in Spriggs
`where there is an interpolation, and then if the block has
`decided that subdivision is needed, it actually generates corner
`addresses and then it puts those corner addresses for the blocks
`that are going to be subdivided onto a stack.
`And that aspect of the coding process reads on the
`analysis system. And then the coding process continues and
`actually talks about how there is a division code 1 that is
`transmitted, and then there is additional corner samples for
`those subdivided regions that are transmitted, and those aspects
`read squarely on the pixel selection which talks about for each
`region.
`
`JUDGE ARBES: So just to make sure I understand
`your position, there is a coding process that's performed by
`processor 14 and certain operations correspond to the analysis
`
`
`
`11
`
`1
`2
`3
`4
`5
`6
`7
`8
`9
`10
`11
`12
`13
`14
`15
`16
`17
`18
`19
`20
`21
`22
`
`
`
`Case IPR2016-00212
`Patent 7,974,339 B2
`
`
`
`system and then certain operations correspond to the pixel
`selection system. Is that right?
`MR. MESSINGER: Yes, correct, Your Honor.
`And I could even go into more detail on sort of how we do that
`mapping, but that's exactly how.
`JUDGE ARBES: And is there enough detail in
`Spriggs to say these are the specific operations that correspond
`to A and these are the specific operations that correspond to B?
`MR. MESSINGER: Yes. In fact, it shows up in
`fig. 6, which is in our petition for the pixel selection system.
`We can go -- well, here is the division, on slide 11, here is
`where it is talking about the actual division of the frame data
`into regions of level of detail which meets the claim language
`of the analysis system.
`And that's an interpolation. And if the values have
`a variance amongst the pixels, then it is subdivided. And
`pixel, as you noted in your Institution Decision, Spriggs has
`extensive teaching on sort of how that subdivision happens.
`And then if you look at fig. 6 this shows, further
`shows how the pixel selection system is carried out in a
`different part of the coding process. Once the divisions have
`been made and those corner addresses of a subdivided region
`
`1
`2
`3
`4
`5
`6
`7
`8
`9
`10
`11
`12
`13
`14
`15
`16
`17
`18
`19
`20
`21
`22
`
`
`
`12
`
`
`
`Case IPR2016-00212
`Patent 7,974,339 B2
`
`
`
`identified, then a division code is sent. And that's in, like, for
`example, row 1 where it says SE, SF, SG, if the interpolation
`shows that there is going to be a subdivision, then the corner
`addresses of those four new blocks that are going to be
`subdivided -- that are subdivided, those are stored on a stack --
`and I will show you fig. 4 which shows that in more detail if
`you like -- but here what happens is those corner addresses go
`in a stack. That's the analysis system.
`And then the process continues and the division
`code 1 on the first row, for example, is transmitted, and then
`the corner samples for those regions that are subdivided, which
`in this case is SE, SF, SG, SI, SH, those are transmitted. Fig.
`4 actually puts it in flow chart form and lays out the different
`aspects.
`
`JUDGE HUDALLA: In your petition did you
`actually mention those aspects of figure 4? I don't recall
`seeing that in there.
`MR. MESSINGER: In our petition we focused on
`fig. 6 because it was a summary. And this summary reads
`better on that broad, general language that just says, in the
`pixel selection, for example, it says receiving region data and
`then generating one set of pixel data for each region forming a
`
`
`
`13
`
`1
`2
`3
`4
`5
`6
`7
`8
`9
`10
`11
`12
`13
`14
`15
`16
`17
`18
`19
`20
`21
`22
`
`
`
`Case IPR2016-00212
`Patent 7,974,339 B2
`
`
`
`new set of data for transmission. This actually shows that
`broad claim language better. So we showed that.
`But it is clear from the overview section of our
`petition in Spriggs, and then the overview of the combination,
`that we are relying on that same algorithm that is in fig. 4.
`And we have additional testimony as well in our reply
`declaration from Dr. Grindon.
`JUDGE HUDALLA: Maybe you could address the
`idea of the pixel data and the region data and whether those
`two aspects are different in Spriggs or is there some common
`data amongst those two things?
`MR. MESSINGER: Yes, Your Honor. In fact, if
`we look at slide 18, with respect to the pixel data, the way we
`refer to it in the petition is we refer to the pixel values, the
`pixels, the pixel data reading on the claim language, and we
`cite that at petition 30, petition page 30, page 31, 44, and 51,
`and then the original filed declaration by Dr. Grindon, 117 to
`118. It is pretty clear that when we refer to pixel data, we are
`referring to the pixel data picture elements in Spriggs.
`For the region data there is language in the petition
`that refers to the corner coordinates for each of the blocks, and
`that corresponds to the region data. And we also identified
`
`1
`2
`3
`4
`5
`6
`7
`8
`9
`10
`11
`12
`13
`14
`15
`16
`17
`18
`19
`20
`21
`22
`
`
`
`14
`
`
`
`Case IPR2016-00212
`Patent 7,974,339 B2
`
`
`
`division codes. And I would just, in particular, we mention it
`throughout, but two pages of the petition that you should really
`highlight are petition page 27 and petition page 30.
`We can look at them if you like but those actually
`sort of reflect that we are using the term region data as the
`coordinates of the blocks as well as the division codes that
`identify it, depending on the part of the process.
`JUDGE HUDALLA: Could I ask you a question
`about what we are looking at right now, I guess, under number
`2 there, the third tick point, where it says "blocks as defined
`by corner coordinates and values."
`Is it your contention that the values themselves are
`not part of the region data?
`MR. MESSINGER: Well, as construed by the
`Board, region data is just the coordinates. In the original
`petition we do refer to the coordinates and values in a more
`general way just because that's the entire block and the block
`is identified by the coordinates, and we mention values, but
`when you look at the entire petition, what we are relying on
`under the Board's construction of region data is the
`coordinates.
`
`1
`2
`3
`4
`5
`6
`7
`8
`9
`10
`11
`12
`13
`14
`15
`16
`17
`18
`19
`20
`21
`
`
`
`15
`
`
`
`Case IPR2016-00212
`Patent 7,974,339 B2
`
`
`
`But when you look at the original analysis
`selection system, it basically says receiving frame data and
`generating region data, and that's the coordinates of a block.
`JUDGE HUDALLA: So you are saying, has your
`position then changed based on our construction in the
`Decision on Institution?
`MR. MESSINGER: Well, I'm saying what is most
`essential is the corner coordinates. The region data, in terms
`of the values, they are actually present in Spriggs and they are
`present in the algorithm in the way it works, because what
`happens is the corner addresses get stored and there is even a
`description we looked at on slide 11 in Spriggs where it says
`one layer of a stack will have the coordinates, one layer -- and
`then one layer will also have the values.
`We were emphasizing it for completeness just to
`show that the region is really identified by the coordinates.
`The values are present but you do not need to rely on the
`values for region data, and the Board was correct in that.
`In fact, you can see bullet point 3, we sort of
`mention in our reply there is even more discussion of how that
`claimed pixel data is not by itself the region data.
`
`1
`2
`3
`4
`5
`6
`7
`8
`9
`10
`11
`12
`13
`14
`15
`16
`17
`18
`19
`20
`21
`
`
`
`16
`
`
`
`Case IPR2016-00212
`Patent 7,974,339 B2
`
`
`
`JUDGE ARBES: Counsel, can you talk a bit about
`the division codes? You also say that the division codes can
`be region data.
`Should we consider the region data to be both then ,
`the division codes and the coordinates, or one or the other?
`What is your position?
`MR. MESSINGER: Yeah, our position is that it
`has to be the coordinates. And in the implementation of
`Spriggs that we are relying on for obviousness, the division
`codes as well. And it is partly because region data is used in
`two elements. So it is used in the analysis to generate the
`region frame, the region data of high and low detail, and that
`can be the corner coordinates.
`And then it also in that step it does generate the
`division coordinates, the division code 1 or zero. And then in
`the pixel selection step it takes that region data, which Patent
`Owner says the pixel selection system is not present, how does
`it do that, but that's completely contrary to the reading of
`Spriggs which says that those coordinates are stored on the
`layer of the stack and then, when there is a subdivision, a
`division code 1, which is properly the -- is transmitted, and
`that reflects the subdivision.
`
`1
`2
`3
`4
`5
`6
`7
`8
`9
`10
`11
`12
`13
`14
`15
`16
`17
`18
`19
`20
`21
`22
`
`
`
`17
`
`
`
`Case IPR2016-00212
`Patent 7,974,339 B2
`
`
`
`That's why we call it part of the region data
`because it is yet another identifier that the receiver uses to
`identify the region. And then it also sends the additional
`corner samples. And that is in the pixel selection system.
`And the way the pixels are selected are based on
`the corner addresses that are identified in that layer of the
`stack. The way the receiver reassembles it is using those
`division codes plus corner -- plus coordinates to reassemble the
`block.
`
`JUDGE ARBES: How is the pixel being selected
`based on the region data? Claim 7, for instance, it says it has
`to be based on it. How is the pixel selected based on that?
`MR. MESSINGER: How is it selected based on?
`JUDGE ARBES: On the region data. So you say
`that the coordinates and the division codes are the region data.
`How is the pixel selected based on that?
`MR. MESSINGER: Well, maybe we could look at
`slide 17. The way it is selected based on that is that once there
`is a division of the frame that has been identified to be
`subdivided, then the coordinate addresses of that are stored on
`a stack, and then the division code 1 is transmitted, and then
`
`
`
`18
`
`1
`2
`3
`4
`5
`6
`7
`8
`9
`10
`11
`12
`13
`14
`15
`16
`17
`18
`19
`20
`21
`
`
`
`Case IPR2016-00212
`Patent 7,974,339 B2
`
`
`
`the actual corner samples for those new subdivided regions are
`transmitted.
`And they are actually looked up, because it has to
`-- to transmit those corner samples it has to use the coordinate
`addresses of those regions. That's how it knows which corner
`samples to send. And maybe you can look back at fig. 6, if
`you want.
`
`It is laid out in fig. 4 as well, but as our petition
`said for pixel selection system, right under the claim language
`pixel selection we put fig. 6. What gets sent in the pixel
`selection system are those corner samples. So how does that
`rely on the region data like you asked? The only way it can
`get those corner samples SE, SF, SG, is looking at those
`corner -- those coordinate addresses that were stored on that
`stack layer, and then you can look at fig. 4 real quick.
`Fig. 4 says it even more clearly in terms of the
`minutia. It is clear that in fig. 6 it is relying on the addresses
`because that's how it finds the samples. It can't find SE, SF
`unless it looks up the address of SE, SF. And fig. 4 actually
`explicitly says that.
`Fig. 4 of Spriggs, and actually scroll down. So if
`you look at fig. 4, this is what I've been describing. It is
`
`
`
`19
`
`1
`2
`3
`4
`5
`6
`7
`8
`9
`10
`11
`12
`13
`14
`15
`16
`17
`18
`19
`20
`21
`22
`
`
`
`Case IPR2016-00212
`Patent 7,974,339 B2
`
`
`
`summarized in fig. 6, fig. 4 has even more details on it, where
`after the interpolation which is above and within the margin,
`there's the -- within the margin is the interpolation step. And
`then once it decides that it has to divide it generates corner
`addresses of the subareas.
`They are pushed onto a stack. And then when it
`transmits the additional corner samples, the way it can do that
`is it is reading those addresses on the stack to sort of send that
`out. I mean, that's kind of how this works, otherwise it
`wouldn't know how to find the corner samples.
`Does that answer your question, Judge Arbes?
`JUDGE ARBES: Yes.
`MR. MESSINGER: And so to answer your
`question as far as construction, our view is that the division
`code is a version of the coordinates in that the system at the
`receiver knows the original coordinates, and then with the right
`division code information it can calculate coordinates. So that
`is what is a little tricky about this.
`If there are no more questions on Spriggs, I will
`move to Belfor in view of Thyagarajan.
`As I mentioned earlier, the essential dispute is
`whether a person of ordinary skill in the art would have
`
`1
`2
`3
`4
`5
`6
`7
`8
`9
`10
`11
`12
`13
`14
`15
`16
`17
`18
`19
`20
`21
`22
`
`
`
`20
`
`
`
`Case IPR2016-00212
`Patent 7,974,339 B2
`
`
`
`combined the non-uniform size blocks of Thyagarajan with the
`uniform size blocks of Belfor. And maybe if we look at slide
`20, I will just touch on how Belfor works.
`Belfor discusses advantages, disadvantages of
`using large block sizes, small block sizes, even discusses an
`ideal case where you sort of segment the regions into
`non-uniform regions. But then Belfor does decide to use a
`single block size. And it calls it an important system
`parameter in the example implementation.
`But for what Belfor does it takes that single block
`size and then it uses different modes and sampling analysis to
`vary the frequency of pixels. So that way if you have an image
`that you want, it has a lot of fine detail, and you want to
`sample more heavily, you can use mode 1 and get every pixel.
`If you have an image that is like, say, blue sky with not as
`much spatial variance, then you could use mode 3 and actively
`represent it.
`In some ways Belfor is even smarter than the '339
`patent in that it also continues and has sort of an extended
`discussion about how you can even allocate those modes, in
`that when you are looking at the different blocks and you want
`
`1
`2
`3
`4
`5
`6
`7
`8
`9
`10
`11
`12
`13
`14
`15
`16
`17
`18
`19
`20
`21
`
`
`
`21
`
`
`
`Case IPR2016-00212
`Patent 7,974,339 B2
`
`
`
`to say do I want to give it a mode 1 or a mode 2 or mode 3, it
`actually has a process that it goes through.
`And it is a pretty simple process. But what it does
`is it basically uses mode 3. It says, well, let's pick the least
`amount of pixels, which would have the lowest quality image,
`and, if we are still under our bit rate, well, then we can go look
`at other blocks and use mode 2 and actually get a little higher
`quality image as long as we are staying under our bit rate.
`JUDGE HUDALLA: Where is that discussion you
`are referencing right now?
`MR. MESSINGER: It is in the mode allocation
`problem that the Patent Owner references. It is in Belfor. I
`think it's on page -- I would have to look at the page to get you
`the cite. The Patent Owner has an extensive discussion of it.
`I only highlight it here because that mode
`allocation problem is kind of one way in which Belfor
`describes how to allocate modes to meet a desired bit rate. It
`does not -- it is not the kind of allocation problem that would
`prevent anyone from combining Thyagarajan's non- uniform
`size blocks.
`In fact, as we pointed out, if you have non- uniform
`size blocks, you can still allocate modes to those non-uniform
`
`1
`2
`3
`4
`5
`6
`7
`8
`9
`10
`11
`12
`13
`14
`15
`16
`17
`18
`19
`20
`21
`22
`
`
`
`22
`
`
`
`Case IPR2016-00212
`Patent 7,974,339 B2
`
`
`
`size blocks. It doesn't preclude someone from combining or
`relying on the non- uniform size blocks of Thyagarajan.
`JUDGE HUDALLA: I have located the mode
`allocation problem section. Does that have any discussion
`about how it is done with non-uniform blocks?
`MR. MESSINGER: No, because Belfor is
`concerned with that example implementation of a common
`single block size when it is talking about the mode allocation
`problem. You are right, Your Honor.
`And if we look at slide 23, just to sort of show how
`the combination of Thyagarajan works so that it would handle
`non-uniform size blocks, what we are relying on is that block
`size assignment 108 highlighted in yellow. That's where the
`subdivision occurs. And fig. 3A teaches the divided image
`block.
`
`And there is actually not that much -- there is not a
`dispute from the Patent Owner, in fact, that Thyagarajan
`teaches divided image block. The dispute comes in terms of
`what information is needed from Thyagarajan to make that
`combination.
`Our position, as laid out in the petition, is that a
`person of ordinary skill in the art, looking at the uniform size
`
`
`
`23
`
`1
`2
`3
`4
`5
`6
`7
`8
`9
`10
`11
`12
`13
`14
`15
`16
`17
`18
`19
`20
`21
`22
`
`
`
`Case IPR2016-00212
`Patent 7,974,339 B2
`
`
`
`blocks, and then choosing to now subdivide to get that
`additional level of detail and balance between big and small
`blocks, which is something Belfor mentioned was a concern,
`would know how to use the PQR data. It would know
`Thyagarajan doesn't teach dividing blocks. It also teaches the
`information you need to divide the block.
`And that, for example, is fig. 3C with the PQR
`data. And so with a person combining it, they would just know
`that in Belfor, you send -- if you are sending over uniform
`blocks, well, then you have one block size. But if you are
`using non- uniform size blocks, well, then you need to send
`different block sizes or information on how the blocks are
`divided.
`
`In some ways the PQR data is kind of like the
`division codes we've talked about. It is the bits that show how
`that block is divided. And a person of ordinary skill in the art,
`it wouldn't preclude them from combining it. They would
`know that there is some additional information that is needed
`because now you have non- uniform size blocks, but it's not
`enough that would prevent the combination.
`And Vedanti's own expert even admitted that the
`amount of bits that you would need to sort of identify
`
`
`
`24
`
`1
`2
`3
`4
`5
`6
`7
`8
`9
`10
`11
`12
`13
`14
`15
`16
`17
`18
`19
`20
`21
`22
`
`
`
`Case IPR2016-00212
`Patent 7,974,339 B2
`
`
`
`non-uniform size blocks is on the order of six bits or less, you
`know, it is not a lot of information.
`So if we look at slide 28, there is the different
`modes. And even though the claims are such a high level, they
`don't require any kind of particular language that would sort of
`be precluded by a mode allocation, a person of ordinary skill in
`the art, even looking at that mode allocation problem, could
`apply it to the non- uniform size blocks. They would just --
`you could go through and pick the mode 3, the lowest sampling
`frequency, and then, if you had enough bit rate, you could up it
`to mode 2.
`Patent Owner says that this comparison would
`involve a lot of work, but this is within the skill level of a
`person of ordinary skill in the art.
`JUDGE HUDALLA: Do you need the discussion of
`your expert in the supplemental declaration to support that
`contention that it is ordinary skill in the art, because I notice
`that you have quite a lengthy discussion?
`MR. MESSINGER: Yes, we do in terms of the
`particular block size information for non- uniform size blocks,
`if you go to that level of detail with the mode allocation
`problem, but all of that was raised by the Patent Owner.
`
`1
`2
`3
`4
`5
`6
`7
`8
`9
`10
`11
`12
`13
`14
`15
`16
`17
`18
`19
`20
`21
`22
`
`
`
`25
`
`
`
`Case IPR2016-00212
`Patent 7,974,339 B2
`
`
`
`What our petition says is what reads on the claims
`themselves. The claims just say generating, receiving region
`data and sort of generating pixel data for transmission. And
`our petition appropriately says that you could combine
`Thyagarajan, which teaches how to subdivide a block, with all
`of the information about how you subdivide the block and have
`bits for that. We rely on that to teach the general -- the
`language needed for the claims.
`When you get down to, well, they couldn't be
`combined because of this particular mode allocation problem,
`that's when we provide additional evidence saying, well, a
`person of ordinary skill in the art wouldn't think that's the kind
`of problem that would prevent a combination. And we
`identified even more information on how it would work for the
`mode allocation with non-uniform size blocks.
`JUDGE ARBES: Counsel, doesn't this argument
`really go to how a person of ordinary skill in the art would
`have meshed the teachings of these two references together?
`Why is that not something that should be in the petition?
`MR. MESSINGER: Your Honor, the petition is
`required to teach why the claims are obvio