[00:00:00] Speaker 06:
Good morning.

[00:00:01] Speaker 06:
We have four argued cases this morning.

[00:00:03] Speaker 06:
The first is number 21, 2085, Unalock, 2017, LLC versus Netflix.

[00:00:11] Speaker 06:
Mr. Cummings.

[00:00:21] Speaker 01:
Good morning, Your Honors.

[00:00:23] Speaker 01:
May it please the Court.

[00:00:25] Speaker 01:
This is a case where the patent trial and appeal board committed reversible error

[00:00:30] Speaker 01:
by construing two patent claim terms without heeding this court's directive from Phillips to construe the terms in the context of the entire patent and such that the construction comports with the instrument as a whole.

[00:00:52] Speaker 01:
This court should reverse the judgment of the board for both of these errors.

[00:01:00] Speaker 01:
With both of these terms, the board ignored passages in the patent that criticized the prior art and explained how the invention operates differently.

[00:01:14] Speaker 00:
Can I just double check something?

[00:01:17] Speaker 00:
So when I read the board's opinion and your papers below in the blue brief, it seemed to me that you didn't have anything compelling in the

[00:01:30] Speaker 00:
to go against what the board had said.

[00:01:34] Speaker 00:
But then in the reply brief, you point to two or three passages that seemed considerably more helpful to you than anything you had cited before.

[00:01:47] Speaker 00:
Should we disregard that new spec support, which you did not present to the board and you didn't even present to us in the blue brief, or should we consider it?

[00:02:01] Speaker 01:
Admittedly, the spec support before the board could have been more in depth.

[00:02:10] Speaker 01:
But the board recognized that we'd argued not by pixel unit, but by a macro block unit.

[00:02:19] Speaker 00:
No, I'm not suggesting that you've changed your position at all.

[00:02:23] Speaker 00:
But the spec support, basically,

[00:02:28] Speaker 00:
bottom of column one, the very top of column two seems... I want to ask you about that support, but I'm just asking you at the initial stage, should we consider it?

[00:02:42] Speaker 00:
Because that was nowhere, unless I missed something, in either your blue brief or in anything you submitted to the board.

[00:02:50] Speaker 01:
Well, yes.

[00:02:51] Speaker 01:
I mean, this review is de novo.

[00:02:54] Speaker 01:
And this court needs to follow Phillips as well and construe these terms in light of the patent as a whole, such that the construction comports with the instrument as a whole, and that those passages

[00:03:15] Speaker 01:
cast light on what the proper construction should be.

[00:03:19] Speaker 06:
So if they were so clear, why didn't you mention them initially?

[00:03:23] Speaker 06:
It sort of suggests that maybe the reading that you're giving to them now is not necessarily the right reading.

[00:03:33] Speaker 01:
No, I think what the issue is, Your Honor, is... Well, I don't think you're answering my question.

[00:03:38] Speaker 06:
If it was so clear from these passages that your position was right, why didn't you rely on them earlier?

[00:03:45] Speaker 01:
Well, I believe that we did rely on them.

[00:03:48] Speaker 01:
The final written decision refers to the emphasis in the patent that it's not to divide by pixel unit, but to divide by a macro block unit.

[00:04:05] Speaker 06:
Where did you rely on these passages in your blue brief?

[00:04:10] Speaker 01:
In the blue brief?

[00:04:12] Speaker 01:
Like I said, I think our discussion of the initial... Well, the answer is you didn't, right?

[00:04:22] Speaker 01:
I don't believe that those were quoted explicitly.

[00:04:27] Speaker 01:
If there was a more general discussion of the preliminary background section of the patent, then we should have included.

[00:04:38] Speaker 01:
But when we got the red brief,

[00:04:41] Speaker 01:
It made statements about the contents of the specification that just aren't true.

[00:04:47] Speaker 01:
There's one lone paragraph that discusses the dividing in the patent.

[00:04:55] Speaker 01:
Well, that's just not true.

[00:04:56] Speaker 06:
Well, Figure 1 itself supports what the board said, right?

[00:04:59] Speaker 06:
I'm sorry?

[00:05:00] Speaker 06:
Figure 1 itself supports what the board said.

[00:05:06] Speaker 06:
Pardon me?

[00:05:06] Speaker 01:
I don't believe that Figure 1 supports the board.

[00:05:09] Speaker 01:
I believe that Figure 1 supports us.

[00:05:10] Speaker 06:
It talks about the division in this flow chart, and it doesn't get to the macro block approach until the fourth step, until the division's complete.

[00:05:27] Speaker 01:
Step 2 talks about, I'm sorry, let me open this up.

[00:05:36] Speaker 01:
So just to make sure, are you talking about the division or the... Yeah, the division.

[00:05:41] Speaker 06:
There's no mention of the macro block approach in connection with division in Figure 1.

[00:05:46] Speaker 06:
It says region division, and then it's not until you get to the fourth step to determine whether the coding process of the pixel value of the macro block is to be executed or not.

[00:05:57] Speaker 01:
Well, the division is done in Block 10.

[00:06:00] Speaker 01:
By the time Block 10 is finished, the object... That's the point.

[00:06:04] Speaker 06:
There's no mention of

[00:06:06] Speaker 06:
utilization of pixels or macro blocks or any particular method for the division?

[00:06:11] Speaker 01:
Because that's in the discussion of the specification.

[00:06:14] Speaker 01:
The specification refers to block 10.

[00:06:20] Speaker 01:
But see, this is a problem that the board had.

[00:06:24] Speaker 01:
They focused on one paragraph in the specification, and they ignored these preliminary passages that explain

[00:06:36] Speaker 01:
that, for example, with division, the prior art divided pixel by pixel.

[00:06:46] Speaker 06:
Well, it's not so surprising that they didn't discuss those passages because you didn't call them to their attention.

[00:06:51] Speaker 01:
We did call to their attention the emphasis in the specification that this invention doesn't divide pixel by pixel because

[00:07:06] Speaker 01:
It's too calm.

[00:07:07] Speaker 00:
It's too... So can you elaborate on that?

[00:07:10] Speaker 00:
How is it that dividing the original input into regions macro block by macro block serves some purpose of this invention in a way that doing the division pixel by pixel and only post division

[00:07:35] Speaker 00:
overlaying a macro block sort of plot on top of it, does it?

[00:07:43] Speaker 01:
Okay, as the specification notes, when you have a system that has low computing power, then it's difficult to perform in real time intensive computing.

[00:07:58] Speaker 01:
And dividing a video frame pixel by pixel

[00:08:05] Speaker 01:
requires significant computing, and let me explain why that is.

[00:08:11] Speaker 01:
If you've got a frame of high definition video, it has 2 million pixels in it.

[00:08:18] Speaker 01:
But if you define that with the 16 by 16 macro blocks, there's only 8,000 macro blocks.

[00:08:25] Speaker 01:
And so it takes significantly more computing power to compare 2 million pixels than it does to compare 8,000

[00:08:35] Speaker 01:
macro blocks.

[00:08:37] Speaker 01:
And so it's able to do the comparison if it's dividing macro block by macro block rather than pixel by pixel because it's got... I don't remember.

[00:08:50] Speaker 00:
Does something in the spec talk about the resource drain on the division process or just the coding process or both?

[00:09:05] Speaker 00:
You just said that more resources are needed for a division process if you do 256 times the number of comparisons than if you otherwise.

[00:09:19] Speaker 00:
And so that suggests there is a resource saving by doing the division on a macro block.

[00:09:25] Speaker 00:
Yes.

[00:09:26] Speaker 00:
Does the spec talk about that or does it only talk about saving resources in

[00:09:34] Speaker 00:
just coding fewer things.

[00:09:36] Speaker 06:
And the coding doesn't take place at the divisional step, right?

[00:09:41] Speaker 01:
The division is done, and once the division is done, then it goes through and does the coding.

[00:09:46] Speaker 01:
The answer is there's no coding at the divisional step.

[00:09:50] Speaker 01:
Right.

[00:09:51] Speaker 01:
The coding comes later.

[00:09:52] Speaker 00:
It's just the comparing.

[00:09:54] Speaker 00:
And my question is, does something in the specs say we'd like to save comparison costs by cutting into a tiny fraction

[00:10:04] Speaker 00:
the number of comparisons we have to make.

[00:10:07] Speaker 01:
Yes, on column 1, 50 through 54 and 2 through 34 talks about the computational efficiency.

[00:10:20] Speaker 01:
And 41 through 48 talks about... And that happens to be one of the passages that I think you didn't cite until your grade review.

[00:10:33] Speaker 01:
I'm not sure.

[00:10:36] Speaker 01:
And then 150 through 54 is the problem about dividing pixel by pixel.

[00:10:45] Speaker 01:
It's hard to perform real time.

[00:10:50] Speaker 01:
Does that answer your question?

[00:10:52] Speaker 01:
It does.

[00:10:52] Speaker 01:
Okay, I see that I'm in my rebuttal time zone.

[00:10:54] Speaker 06:
Did you have an expert who explained the view that you're now articulating about these passages in the specification?

[00:11:04] Speaker 01:
No, we did not.

[00:11:06] Speaker 06:
So you're just, this is lawyer or you're asking us to make a de novo interpretation of these passages?

[00:11:14] Speaker 01:
We believe that these passages are clear enough that they don't need an expert.

[00:11:21] Speaker 01:
And Dr. Saber's testimony should be disregarded because it's not only inconsistent with the specification, but his own testimony is internally inconsistent, as we noted in the briefing.

[00:11:39] Speaker 01:
I'll reserve the rest of my time for both.

[00:11:43] Speaker 06:
Okay, Mr. McKeever.

[00:11:45] Speaker 03:
Good morning, Your Honors.

[00:11:46] Speaker 03:
May it please the Court, Patrick McKeever on behalf of Appellee Netflix.

[00:11:51] Speaker 03:
With me at council table is my colleague, Dan Bagatell.

[00:11:56] Speaker 03:
I wanted to pick up on a couple of things that Mr. Cummings said and some of the questions that the Court has asked, which I think keyed in on some of the issues here.

[00:12:06] Speaker 03:
This argument that is the thrust of the gray brief about computational efficiency is a completely new argument, as the Court, I think, recognized.

[00:12:16] Speaker 00:
Right.

[00:12:16] Speaker 00:
I mean, OK.

[00:12:17] Speaker 00:
I'm prepared to assume that.

[00:12:19] Speaker 00:
That was certainly my impression.

[00:12:21] Speaker 00:
We had certainly said an interactive gift expressed if they had made it in the blue brief.

[00:12:25] Speaker 00:
Because it's just a new spec material in support of the same claim construction, there would be no forfeiture.

[00:12:32] Speaker 00:
There's a sentence in Interactive Gift Express that says that expressly.

[00:12:36] Speaker 00:
They didn't make this point in the blue brief.

[00:12:39] Speaker 00:
So I'm interested in the question of forfeiture, as to which there probably is some discretion.

[00:12:44] Speaker 00:
And assuming that there's some discretion about that question,

[00:12:49] Speaker 00:
it would matter to me to understand whether this material, including the part at the bottom of column one that was discussed about the resource, the need for resources on the comparison process, not just the coding process, and the material at the top of column two, why that doesn't support the other side very, very clearly.

[00:13:12] Speaker 00:
Because if it's very, very clear, I'm not as inclined to say it's forfeited.

[00:13:17] Speaker 03:
Sure.

[00:13:18] Speaker 03:
I understand, Your Honor.

[00:13:19] Speaker 03:
Thank you.

[00:13:21] Speaker 03:
So it does not.

[00:13:23] Speaker 03:
I mean, one other point is, again, if these arguments had been raised below, you also just asked about whether they had an expert to address these points.

[00:13:32] Speaker 00:
We often don't have experts in claim construction.

[00:13:34] Speaker 00:
It's a matter of law.

[00:13:35] Speaker 03:
Right.

[00:13:35] Speaker 03:
Mostly.

[00:13:36] Speaker 00:
So tell me why these passages are not very clearly in support of your answer.

[00:13:43] Speaker 03:
Sure.

[00:13:44] Speaker 03:
So first of all, there was a question I think Your Honor specifically asked about, does the specification specifically say that the dividing step in the invention does a fewer number of comparisons by doing it at the macro block level than at the pixel level?

[00:14:02] Speaker 03:
It does not.

[00:14:03] Speaker 03:
Mr. Cummings pointed to a couple of passages from the specification that talk about, you know, at a high level how the invention might enable real-time encoding on a low-performance system.

[00:14:16] Speaker 03:
Again, nothing in the claims about real-time, nothing in the claims about low-performance systems.

[00:14:21] Speaker 00:
But there is no statement in here that says that by virtue of doing the dividing – Well, let me just – so let me just do this sentence, which is column one, line 52, 54.

[00:14:32] Speaker 00:
Since very complicated processes should be executed in order to divide, this is not to code, in order to divide the video by the pixel unit in the case of using the object-based coding technique utilizing the conventional region division technique, it is difficult to embody such processes in real time.

[00:14:52] Speaker 00:
That seems to me to say dividing, which means comparing, whether it's same or different from previous frame,

[00:15:01] Speaker 00:
is resource intensive and we would like to cut down on the use of those resources in the division process before you get to coding.

[00:15:09] Speaker 03:
Right.

[00:15:10] Speaker 03:
But one of the questions is where exactly in the division process?

[00:15:13] Speaker 03:
The patent describes the division process in column one in this

[00:15:17] Speaker 03:
prior art known technique at a pretty high level, right?

[00:15:20] Speaker 03:
It says you do the frame differencing between the current frame and the previous frame.

[00:15:23] Speaker 03:
Well, when they describe the division process and the invention, they also say it relies on frame differencing between the current frame and the previous frame.

[00:15:31] Speaker 03:
There's not actually a single sentence anywhere in the specification, which is fairly short, that says that

[00:15:38] Speaker 03:
There's a reduced number of comparisons that's performed when you do this at the macro block level.

[00:15:42] Speaker 03:
That's the attorney argument that Unilock is making.

[00:15:45] Speaker 00:
Why isn't that self-evidently true?

[00:15:48] Speaker 00:
Because there's fewer items you're comparing by a fraction of 1 over 256, don't you?

[00:15:54] Speaker 03:
But how do you perform, how do you compare macro blocks, how do you compare brightness of macro blocks without comparing the brightness of the pixels that make up the macro block?

[00:16:05] Speaker 03:
There's no description in the patent specification of how you do that.

[00:16:09] Speaker 04:
And if you look at – That sounds like a written description problem, but not necessarily a claim construction problem.

[00:16:16] Speaker 04:
I don't necessarily disagree with that, Your Honor, but – I guess what I'm struggling with here is if their view of how their invention works is correct, it does seem like the prior art does it in a slightly different way.

[00:16:32] Speaker 04:
And it's very hard.

[00:16:33] Speaker 04:
I found this patent very unclear at how it's written.

[00:16:37] Speaker 04:
And not very detailed either.

[00:16:39] Speaker 04:
But my understanding is the first step before any of this happens is you have to look at everything pixel by pixel with the first image.

[00:16:50] Speaker 04:
And then they say they divide that into macro blocks.

[00:16:54] Speaker 04:
And then when they do the comparison step for the object and the background coding, they do that comparison on a macro by macro block basis.

[00:17:02] Speaker 04:
If that's true, then isn't that different from the prior art?

[00:17:06] Speaker 04:
And the board is wrong.

[00:17:08] Speaker 03:
So it is different.

[00:17:09] Speaker 03:
I would grant you it.

[00:17:10] Speaker 03:
The preferred embodiment of the dividing step in the patent specification is different from the prior art.

[00:17:16] Speaker 03:
And remember, the board didn't invalidate claim two, which is the claim that actually is the dependent claim that tracks the preferred embodiment of the dividing step.

[00:17:27] Speaker 03:
Why didn't the board invalidate it?

[00:17:29] Speaker 03:
Well, that claim was never asserted in the litigation.

[00:17:31] Speaker 03:
It wasn't challenged.

[00:17:32] Speaker 03:
It wasn't an issue in the IPR.

[00:17:35] Speaker 03:
But to pick up, Judge Hughes, on what you just said, I think you're right that even in the dividing step in the invention, and the claim language actually kind of gives this away, you are still doing it.

[00:17:46] Speaker 00:
Just to be quick, claim two is not different only in this respect.

[00:17:51] Speaker 00:
It's narrower yet by comparing test statistics.

[00:17:54] Speaker 00:
So it doesn't quite make your point.

[00:17:56] Speaker 03:
Yes, Your Honor.

[00:17:57] Speaker 03:
That's fair.

[00:17:59] Speaker 03:
But I would say Claim 2 clearly tracks the preferred embodiment of the dividing step in the specification.

[00:18:05] Speaker 00:
And then makes it more specific.

[00:18:06] Speaker 03:
Exactly.

[00:18:07] Speaker 03:
And you know, whereas the dividing step in Claim 1 is recited quite broadly, very similar to how the dividing step is introduced in the past specification.

[00:18:18] Speaker 00:
I just went on a detail issue.

[00:18:19] Speaker 04:
Yeah.

[00:18:21] Speaker 04:
In answering my question, what I'm curious about is actually the specific language in 1a, which again, there might be written description problems with this, but it says dividing the stationary background and the object in a micro block by macro block basis.

[00:18:40] Speaker 04:
So it seems to tell me that it's using the macro blocks to divide up the background and the object.

[00:18:47] Speaker 04:
how they do it can go.

[00:18:49] Speaker 04:
But assuming we take that as a description, why isn't the board's construction inconsistent with this?

[00:18:57] Speaker 03:
So our position, which is consistent with what the board held, is that... Because the board doesn't require this to be done first.

[00:19:03] Speaker 04:
It just requires the results of everything to do this.

[00:19:06] Speaker 04:
But this seems to suggest that this is a step that requires that dividing and

[00:19:13] Speaker 04:
of the two different regions on a macro block by macro block basis.

[00:19:18] Speaker 03:
And so – so, Your Honor, that's not how I read it.

[00:19:22] Speaker 03:
And again, I agree.

[00:19:23] Speaker 04:
I mean, I think that there are many things that are – It's not a very well-written claim, to be sure.

[00:19:27] Speaker 03:
Right.

[00:19:28] Speaker 03:
Right.

[00:19:29] Speaker 03:
I read this language as what this is talking about, consistent with the specification, is this issue of granularity.

[00:19:35] Speaker 03:
Are we trying to create these very detailed

[00:19:38] Speaker 03:
regions that are, you know, have these deviations at the pixel level, or are we trying to do this very blocky, simplistic definition of regions?

[00:19:49] Speaker 03:
Why do we want to do it blocky and simple in this patent is because if you do it that way,

[00:19:54] Speaker 03:
it drastically reduces the amount of data that you have to provide about the shapes of the different regions.

[00:20:01] Speaker 03:
So if you have simple shapes, you don't have to spend a lot of bits conveying what those shapes look like.

[00:20:06] Speaker 04:
Well, and can't you – I mean, I understand it doesn't tell you how to do it, but I mean, theoretically, it seems to me you could just look at the whole map for a block and compare it to the previous one, which is what you do from the previous frame.

[00:20:18] Speaker 04:
And if there's no changes at all, then you keep it.

[00:20:22] Speaker 04:
Because it's part of the background and it's easier to code you don't do extensive coding and if there are changes in the object then

[00:20:32] Speaker 04:
you coded in a different pixel by pixel basis.

[00:20:36] Speaker 04:
And that seems to be doing the dividing and comparing steps on the macro block by macro block basis, which is what they're arguing for.

[00:20:44] Speaker 03:
But you're still looking at it on a pixel by pixel basis.

[00:20:47] Speaker 03:
So the idea that in terms of doing of doing that comparison, and I didn't really respond to your previous question, which

[00:20:54] Speaker 03:
Which is – where do we see which way that goes in the specification?

[00:20:57] Speaker 03:
And I think, again, the spec is very sparse, but if you look at Claim 2, it tells us a little bit more about what this test statistic is that's used in the preferred embodiment of the divided set.

[00:21:11] Speaker 04:
Can I just – I'll let you go on, but what you're essentially saying is even if you do do this

[00:21:16] Speaker 04:
into macro blocks.

[00:21:18] Speaker 04:
You can't really just look at the macro block as a whole to determine whether it's changed.

[00:21:22] Speaker 04:
You still have to look at it on a pixel by pixel basis.

[00:21:25] Speaker 04:
Exactly.

[00:21:25] Speaker 04:
But I think their argument is, no, that's not what we do.

[00:21:29] Speaker 04:
How they do it, again, not explained to me sufficiently.

[00:21:34] Speaker 04:
But let's just assume you can do it on a macro block by macro block basis without looking at the changes in individual pixels.

[00:21:42] Speaker 04:
Isn't that what this is claiming?

[00:21:44] Speaker 03:
I don't believe so because the – at least in Claim 1, Your Honor, you know, again, our position is that the – in a macro block – by macro block basis is talking about the level of granularity at which you are dividing these regions out of the inputted video.

[00:22:01] Speaker 03:
So it's a result.

[00:22:02] Speaker 05:
For purposes of the later coding.

[00:22:04] Speaker 03:
I'm sorry.

[00:22:05] Speaker 05:
For purposes of the coding.

[00:22:08] Speaker 03:
That's right.

[00:22:08] Speaker 03:
I mean, that's the reason that you're doing this at a macro block level.

[00:22:12] Speaker 03:
And the prior art recognizes that as well.

[00:22:14] Speaker 03:
One is you're trying to simplify the amount of data.

[00:22:19] Speaker 03:
You're trying to make it more efficient.

[00:22:20] Speaker 03:
But you're going to end up encoding everything in macro blocks anyway.

[00:22:24] Speaker 00:
So why not get to that point?

[00:22:27] Speaker 00:
What I was asking about before is that there may be two resource intensive processes.

[00:22:33] Speaker 00:
One is the division process, and the other is the coding process.

[00:22:37] Speaker 00:
If doing the division macro block by macro block instead of pixel by pixel saves resources for process number one, that is in service of the goal of this invention.

[00:22:57] Speaker 00:
It doesn't have to be only saving resources for coding, which it would also do if you're coding fewer things.

[00:23:06] Speaker 00:
Right.

[00:23:07] Speaker 00:
claim language be any different if, instead of using the seemingly just incorrect English expression in a basis, it said on a basis?

[00:23:17] Speaker 03:
So I've thought about that question, Your Honor.

[00:23:20] Speaker 00:
Yeah, I assume everybody had the same.

[00:23:22] Speaker 00:
This is not normal English, so it's hard to know what to make of it.

[00:23:27] Speaker 03:
Exactly, exactly.

[00:23:28] Speaker 03:
And that's why, again, I think

[00:23:30] Speaker 03:
And I will respond to that question fully.

[00:23:32] Speaker 03:
But I think you have to look at this and compare it to the specification.

[00:23:36] Speaker 03:
And again, claim one broadly recites the dividing step just as it's introduced in the specification.

[00:23:42] Speaker 03:
And then in our view, claim two is the dependent claim that has the specific implementation of using test statistics and all that.

[00:23:50] Speaker 03:
I don't think it would change my interpretation or Netflix's interpretation of the claim language if it said on.

[00:23:56] Speaker 03:
Because I think you have to look at where does that phrase appear in the claim language.

[00:24:00] Speaker 03:
It doesn't say, and I'm not saying it would change my position necessarily even if you move the phrase, but it doesn't say dividing by taking each macro block of the current frame

[00:24:14] Speaker 03:
or something like that.

[00:24:15] Speaker 03:
It says dividing the stationary background region and the object region from an inputted video.

[00:24:22] Speaker 03:
So it's telling us what we're trying to, the result, right?

[00:24:25] Speaker 03:
The result is we want to create these.

[00:24:27] Speaker 00:
That's why some of the newly cited spec passages seem helpful in figuring out what this means, like the one at column two, line five, which specifically talks about dividing not by pixel unit, but by

[00:24:43] Speaker 03:
Right.

[00:24:44] Speaker 00:
16 by 16.

[00:24:45] Speaker 00:
Right.

[00:24:46] Speaker 00:
And it tends to point in their direction, maybe even more than tends to.

[00:24:51] Speaker 03:
I mean, I disagree.

[00:24:53] Speaker 03:
That's not how I read that portion of the specification.

[00:24:55] Speaker 03:
I think, again, that portion as well is not getting at the unit of processing that you are breaking this thing into for purposes of dividing.

[00:25:06] Speaker 03:
What it's getting at is, what's the granularity we want to get to?

[00:25:09] Speaker 03:
When we create these regions.

[00:25:11] Speaker 03:
For coding purposes, you mean?

[00:25:13] Speaker 03:
Yes.

[00:25:13] Speaker 03:
when we create these regions, do we want them to be, again, detailed, very complex shapes, or do we want these simple block, very blocky shaped regions that are, maybe you could do it better, but it's not worth the processing trade-off, so let's keep it simple and save some bits and just make it really blocky and it's good enough for our purposes.

[00:25:37] Speaker 04:
I'm still struggling because it does seem to me like there are two different

[00:25:41] Speaker 04:
potential methods described here in the prior art, it divides up the object and the background on more of a shape basis.

[00:25:49] Speaker 04:
It like kind of does a broader outline of the object and then everything else is background.

[00:25:57] Speaker 04:
And that's done on a pixel basis, I think.

[00:26:01] Speaker 04:
And what they're trying to do is pre-divide the whole frame into macro blocks and then divvy up the macro blocks into object or background.

[00:26:11] Speaker 04:
and it's rougher because it's not a pixel by pixel basis, but that does seem, if that's what the claim is, to be different from the prior art.

[00:26:19] Speaker 03:
I mean, if the dividing step in claim one is describing the result, which is our position, then the result is the same.

[00:26:29] Speaker 04:
No, but that's not their position, and that's not the way they read the claim language.

[00:26:34] Speaker 04:
And it's, again, very badly written.

[00:26:38] Speaker 04:
sentence.

[00:26:39] Speaker 04:
But if you read it as it's actually at the beginning stage where the very first frame has to be divided up.

[00:26:47] Speaker 04:
I understand.

[00:26:47] Speaker 04:
I'm going to finish quickly.

[00:26:48] Speaker 04:
But once you do the comparisons of the first macro block to the second macro block frame, they're divvying up isn't on a pixel by pixel basis about what's object and background.

[00:27:00] Speaker 04:
It's just

[00:27:01] Speaker 04:
This whole block has nothing but background.

[00:27:03] Speaker 04:
This whole block has part background, part object, maybe all object.

[00:27:08] Speaker 04:
And so they're doing that.

[00:27:09] Speaker 04:
And if that's the case and the proper reading of that claim, hasn't the board erred?

[00:27:14] Speaker 03:
I would say no, because that is the same result that you get with Tellori's method.

[00:27:20] Speaker 03:
And that's why Tellori does this tiling with macro blocks.

[00:27:23] Speaker 03:
They're trying to get to the same result.

[00:27:26] Speaker 03:
It's a more sophisticated algorithm that they use to get there.

[00:27:30] Speaker 03:
It's more refined.

[00:27:31] Speaker 03:
But ultimately, they're trying to, they do this, you know, pixel level process, but then they... So you're saying this, even if we read this claim language that way, it's just claiming the intended result for purposes of coding.

[00:27:45] Speaker 03:
That's correct.

[00:27:45] Speaker 03:
That's correct, Your Honor.

[00:27:47] Speaker 03:
Thank you, Your Honor.

[00:27:59] Speaker 01:
Just a couple of items, Your Honor.

[00:28:01] Speaker 01:
I do want to point out that Tulluri has nothing to do with computational efficiency.

[00:28:09] Speaker 01:
Tulluri is entirely directed towards coding efficiency.

[00:28:13] Speaker 01:
It assumes that it has the computational power to do a bit by bit compare.

[00:28:20] Speaker 06:
Now, responding to his... The problem is what's not clear is whether... This patent is certainly talking about coding efficiency.

[00:28:31] Speaker 06:
Yes.

[00:28:31] Speaker 06:
And the coding efficiency feature is shown in the prior art, correct?

[00:28:36] Speaker 01:
As well as the division pixel by pixel.

[00:28:39] Speaker 01:
Yeah.

[00:28:40] Speaker 06:
Well, right.

[00:28:41] Speaker 06:
And the question is, was this

[00:28:43] Speaker 06:
invention designed to save computational power at the division step or the coding step only, right?

[00:28:53] Speaker 01:
We think that the computational savings is directly most at the division.

[00:29:02] Speaker 06:
Most at the division?

[00:29:04] Speaker 01:
I mean, there is some computational savings in the coding if it's done

[00:29:11] Speaker 01:
a macro block by macro block.

[00:29:13] Speaker 01:
But the patent emphasizes that the primary savings by dividing by macro block is computational.

[00:29:22] Speaker 06:
And then you get... Well, that's a computational saving in the coding step.

[00:29:28] Speaker 06:
I mean, you are saving computing power, whatever, at the coding step, right?

[00:29:36] Speaker 01:
There is some savings of the coding step, but the patent's emphasis on the computational savings is on the division step.

[00:29:44] Speaker 01:
And I want to respond to one thing.

[00:29:46] Speaker 06:
Where exactly do we find that?

[00:29:48] Speaker 01:
It's column one, line 50.

[00:29:58] Speaker 01:
Since very complicated process,

[00:30:01] Speaker 01:
should be executed in order to divide a video by pixel unit.

[00:30:05] Speaker 01:
Is that paragraph there where it talks about dividing by pixel units is very computationally intensive?

[00:30:15] Speaker 01:
Now, responding to one thing that he mentioned, that dividing by macro blocks is you have to do pixels too.

[00:30:24] Speaker 01:
That's not what Dr. Saber testified.

[00:30:29] Speaker 01:
in his declaration, appendix 529 through 532, in his background section, he talked about and gave an example.

[00:30:41] Speaker 01:
It was well known that pixel-level segmentation was computationally intensive.

[00:30:46] Speaker 01:
He showed an example using 4 by 4 macro blocks to avoid the detailed pixel-level segmentation.

[00:30:55] Speaker 01:
This is where his declaration

[00:30:58] Speaker 01:
was inconsistent.

[00:31:00] Speaker 01:
He recognized that macro block division saved.

[00:31:05] Speaker 01:
But then when it came time to apply the claim limitation, his opinion was, well, it doesn't matter if you divide pixel by pixel so long as the result is macro blocks.

[00:31:20] Speaker 00:
And at least theoretically, I think I understood your friend on the other side to make a suggestion, which I was translating in my mind in the following way.

[00:31:28] Speaker 00:
If you have 256 pixels, you might have 256 luminance or chrominance.

[00:31:36] Speaker 00:
Is that the right word?

[00:31:37] Speaker 00:
Yes, chrominance measures.

[00:31:39] Speaker 00:
And then you need to figure out, well, what is the comparison if you take the whole block at a time?

[00:31:44] Speaker 00:
So you need to figure out what you're somehow going to aggregate the different luminances and the chrominances, or if any single one of them is different.

[00:31:53] Speaker 00:
And that's nowhere discussed in this fact.

[00:31:56] Speaker 01:
It's nowhere discussed in this spec, because it's assuming.

[00:32:00] Speaker 00:
If you had to do all of that, then perhaps there might not be any computational savings.

[00:32:07] Speaker 01:
Not according to the disclosure that's in this spec.

[00:32:10] Speaker 01:
Which then raises the written description question.

[00:32:12] Speaker 01:
I see that my time is up.

[00:32:13] Speaker 01:
Thank you, Your Honor.

[00:32:15] Speaker 06:
Any further questions?

[00:32:16] Speaker 06:
All right, thank you.

[00:32:17] Speaker 06:
I thank both counsels.

[00:32:18] Speaker 06:
The case is submitted.

[00:32:21] Speaker 06:
Our next case is number 2210.