0:00:13thank you mister chairman and action
0:00:15yeah that into our main easement them together with those of another
0:00:21i will present this way
0:00:22we had integrating this adjustment bishop
0:00:26the algorithm and the G D C code
0:00:30the two used
0:00:33we have a presentation
0:00:35we we introduce you the as as the
0:00:39and uh together with a card does also have a main goal of one main object
0:00:43the we're
0:00:44yeah i move to describe
0:00:47that the contribution which are mainly what in in the integration of a
0:00:51percent of but we him members the system
0:00:53the and the concept of just be should distortion
0:00:57in the code that this kind of to see that for the high efficiency video coding standardization problem
0:01:03and
0:01:03i was also show that was at that you have a from these integration
0:01:08and i conclude my presentation shown shown a some future research that action that can be so you know that
0:01:14to improve
0:01:15a a of the present work
0:01:17let's move to the contents and that
0:01:19but to improve that is where we considered the increasing present an increasing route all
0:01:24hi we finish and and that a definition you using that multimedia applications
0:01:30is is using increase input we have is basically you the recent advances
0:01:34you you catch in this place that not
0:01:37this kind of
0:01:38nation
0:01:39oh you use
0:01:40uh it is back to that of the
0:01:43uh and then we and the the road
0:01:46you know that the pressure of the contact and force that
0:01:49and the know to do not
0:01:51see the the the physical limitation of the net force
0:01:54the compression efficiency must be improved the on the one about the what we consider see what they a state
0:01:59of you are can be you coding which is the each to just T V C step
0:02:04and he's emerging a has to but in general two thousand and
0:02:08a our proposed between an that in the high compression all
0:02:14in math
0:02:15uh two thousand and twenty seven proposals submitted
0:02:21for marketing and histories
0:02:23and set the proponents that were able to provide that
0:02:28two shows the same subject
0:02:31you could be doing this
0:02:33we first and that
0:02:34but only
0:02:35how the way
0:02:37is it
0:02:38and
0:02:40seven first about additional they do you see it can you
0:02:46which is true
0:02:47um
0:02:48i have a a a a us that as a nation of search people use a limit to the high
0:02:52efficiency did you could
0:02:54that it is that it is the in regards the integration of the most polite putting tunes proposed you the
0:03:00core was that
0:03:02in a a a a a a a a a which
0:03:05time time this application was code has an not that are close to the relation of things to G be
0:03:09see
0:03:10uh
0:03:11right and small that right but that's H and i two is basically
0:03:16a a and a kid that of for a a motion compensated for be the you those
0:03:22but has seven know that is
0:03:24most most part about in the block size is this size uses for the of
0:03:29and that are shown how about
0:03:31but not the uh
0:03:34or another
0:03:35he a filter but that the we in feed that which use of you know you
0:03:40but has a a a a a a uh scroll of just that can to improve the compression efficiency
0:03:48that was so that a is that were i am and businesses it incorporates the
0:03:52and that that this is because i well you high is not you from perceived
0:03:57the distortion introduced by quantisation and most coding
0:04:00but to simplify a my and to have the be a that is being
0:04:04"'cause" so one times of maybe the project doesn't properly and of the distortion but
0:04:09see that the the proposition that was you or over the web age
0:04:14have some as well the distortion use the last issue in particular that as a round the line not have
0:04:20and i that's we had a discourse used by opening but you
0:04:24we particular that in the in space a sure than the you going to the ground
0:04:28how how you may be the system provides a actually
0:04:31a a a a a a space distortion sensitive he map and use my and we can X
0:04:37in order to perform a also the quantisation what the human visual system
0:04:41is less sensitive sitting
0:04:42major major is but you want to systems that sensitive that are of quantization of Y
0:04:49but money is that
0:04:50is that a as it did easy use is usually that so a little by means of the concept of
0:04:56just will store
0:04:58which are was to the mean is a true should be the we should no change can be a C
0:05:02but you human be seat
0:05:05but is that this trash which is
0:05:07no computed by means of a just a but was partially
0:05:11we can are set of your perception the late the initial positions that you know to obtain a perceptual quantization
0:05:18is that she julie and B
0:05:20the from this
0:05:21basic intuition from do the basic idea
0:05:24can't that to work which is which consist in the integration of a just a visual distortion my
0:05:30the H E C code can in particular team
0:05:33address
0:05:34well X
0:05:35i
0:05:35are needed
0:05:36you know to account for the no it really be used that one see are in the each E P
0:05:40C's that is H
0:05:41pro
0:05:42so let's move
0:05:43a see are these
0:05:44stationary stationary char
0:05:46they
0:05:47a a lot more
0:05:48uh actually a technical contribution but before i like to a that some you the J E but that we
0:05:55are used for these work
0:05:57it is a state of the uh an you know
0:05:59which were
0:06:00in the transform domain actually provide
0:06:03they do a traction and more transform me
0:06:06a well one position or money
0:06:07takes place in a you couldn't stand
0:06:10and uh i'm not really a trash or for the D C
0:06:13for the frequency band a yeah in the key in H
0:06:16you don't has the multiplication of a
0:06:18to to its that takes into account
0:06:21different spatial masking make it is not man i
0:06:24but the first uh
0:06:26takes into account the fact that you want a shows
0:06:29a you are the works expect
0:06:31a frequency band
0:06:32seven and take into account the fact how many high less sensitive
0:06:37distortion
0:06:38in to those in bright and dark a major areas
0:06:41but particular collect you you and the
0:06:43the first frame of the to the sequence
0:06:45so that the part of as
0:06:47but that that that uh that is that the has not shoes that S on the but in the background
0:06:52the will see that higher to an attraction
0:06:55and in particular the last to takes into account the fact that that that is that be giving the you
0:07:01i is not sensitive to
0:07:02distortion introduced in particular
0:07:04and that didn't than new by major at her sides
0:07:08text to that get a major
0:07:11so
0:07:12this is this
0:07:12but of the state that you had in G D more you but used a lot
0:07:17see is
0:07:18as a but size it L eight block size and working point to
0:07:22a a discrete cosine transform a
0:07:24are may take a um has a frequency decomposition composition a frequency transformation
0:07:30the
0:07:30T D C
0:07:31the are the but he's there are
0:07:33the the as i told you but you as in the web
0:07:37and uh they are the shown or or a and not a frequency decomposition
0:07:41this discrete "'cause" and transform a beach
0:07:43use the more that and the directional transform a
0:07:46but but we to put the the goals
0:07:50so
0:07:51technicians must be designed in order to integrate the these in the more the each you C clear that and
0:07:56i
0:07:57stands they may have a bad
0:07:58first to more than a
0:08:00in particular the terms that uh actually more robust that
0:08:03uh uh that didn't you might high mass in related to the frequency band
0:08:09i J
0:08:10so that's so which uh uh which are using
0:08:12extensions
0:08:13and the a list that's i shows look uh related to the into a job just because an transform manner
0:08:19to to use of higher size
0:08:21and
0:08:22we observe that for
0:08:24and then uh
0:08:26is use in the G P C code is the same as a a a i'm using the H is
0:08:31for
0:08:32for is kind of a
0:08:34those is
0:08:35i i i and you know the which is "'cause" that you to buy the the perception but in matches
0:08:40is provided with a
0:08:41so
0:08:42as as me that there was a these is is that uh
0:08:45a a computer um
0:08:47actually actually a by means of a subjective experiments that the for the apply
0:08:52apply design for a spot
0:08:55oh
0:08:56we also have
0:08:57how are and they hate okay we have see the perceptual metric but is i a link which is that
0:09:04and
0:09:04perceptual metric that provides an estimate of the perceptual quality of
0:09:08perceived quality
0:09:09and the system
0:09:11i
0:09:12uh uh and um no combine the quantization error
0:09:17by means of the J the corresponding to you
0:09:21the for meant to provide the per enjoy and you
0:09:25and for this kind of a matrix
0:09:27and in particular
0:09:28the be ct out four
0:09:31but side is
0:09:32a i i and you but some care of a subjective experiments in order to
0:09:36a a a a a a a a a a a relationship between that the frequency which is basically the
0:09:41way is that you have a state of each
0:09:44high frequency band
0:09:46and then more positions that used to obtain a dense willing
0:09:53uh matrix back
0:09:54and this change
0:09:56so it's for the metric
0:09:59but by definition
0:10:01putting in the initial G B
0:10:03that's positions that she is that
0:10:06but just the should a special trash of the for
0:10:09by inverting the of relation here
0:10:11we yeah this tension
0:10:13a the J D bond or my
0:10:15i use and
0:10:16in the H in the mobile into to collecting the each E C code
0:10:20for a a a a a special that it to the mall but and the transfer has to do that
0:10:25actually that
0:10:27a a uh and image transfer basically it is the projection of a good date thing or enough for about
0:10:33see
0:10:34well as which is uh an system was can see you the but that would result that the back of
0:10:38frequencies you
0:10:39the case for the creek as an transfer
0:10:42well for the kind of but but then the directional transform of these is that
0:10:46a reference is rotated
0:10:49i the to the prediction that were actually use for code
0:10:53for this you will have a a a a um
0:10:56the that's i shown that list of the uh since it to the T for the frequency band but you
0:11:01want high
0:11:02we to use that and use that the can is
0:11:06you notation of the J D is cube and two
0:11:09according to the angle at which is that the handle you
0:11:13for inter prediction direction it she C
0:11:17so you you was that
0:11:19this was that it
0:11:19stations and not like to show you some uh a a a a little was out that we have obtained
0:11:25it a yeah uh and that of this may then we have a a a remote particular interest in a
0:11:33high definition you can't and
0:11:35the for have used to say
0:11:37it did contents used but the J C T V C
0:11:40and that they couldn't crowd
0:11:42a that is a a a a a G D C code it use so
0:11:46has been so that that but do this
0:11:48shouldn't be by the joint whatever
0:11:50you know you go
0:11:51but for
0:11:52that that are also yeah have show that was lot
0:11:55a present that to but each of these two clusters
0:11:59right
0:12:00feature space is this a that an S
0:12:04the index
0:12:05where as these indices there were have been computed according to the high used
0:12:09iteration
0:12:13got in the assessment of to you we have to use it as a distortion measure the multiscale is that
0:12:18a similarity metric which is a a a a matter which is an extension for if
0:12:23a of that better well no
0:12:24a a similar to magic proposed but
0:12:27a a professor at all
0:12:30and uh are saying that you can find a details so
0:12:33to and estimate the but is that we have a
0:12:35the design
0:12:36and is based on the cost of a or sort them by
0:12:40and is
0:12:42uh at the biology actually was to find a process
0:12:46T
0:12:47a perceptually that
0:12:48uh a common point to between the be you know that it's being coded with a G D C code
0:12:54and did you get that is being coded bit a a C that
0:12:57we the J more to great
0:12:59that for to be a perceptual
0:13:01is the same we have able to measure
0:13:03but bit of that actions and i show to the bit right of action between a
0:13:07the J E U my
0:13:09and the to my i H G D C code without out the integration of the G in the model
0:13:16but that actions we have an average
0:13:18but the to the right of actions and in particular
0:13:22uh i as you can or the result
0:13:24i i try and uh well as the bit of the actions were expected it
0:13:28uh uh no perceptual a my
0:13:31it the
0:13:32as where X to the quantization parameters
0:13:35i
0:13:35for
0:13:36hi
0:13:37quantization position parameters you
0:13:39the uh
0:13:40see coefficient efficient a times to zero or got less the duration performed by the jury in track
0:13:48but but that was so we show that that the one to the actual for the but
0:13:52to have a sequence
0:13:53has this is a sequence which can is that i
0:13:56not
0:13:57a a just use me uh right
0:14:00that's but any major and as and the for our team that is able to actually provide a positions that
0:14:06as well i i have a high as
0:14:08how you high was not going to recognise and distortion for these kind of a
0:14:14you so you are
0:14:15same
0:14:15they are organising there
0:14:17yeah so
0:14:18rate distortion of for the two
0:14:20has to be C
0:14:24so a to the
0:14:25conclusion
0:14:26uh here we propose an integration of a a a a a special to email in the
0:14:32called that this country can see that in they each E P is not the position
0:14:37and and use of these good are for sure the
0:14:39that
0:14:40we can do
0:14:41in then
0:14:42improve the compression efficiency because we can going through the actions that the same perception
0:14:47you are they
0:14:48each each D C known perceptual called
0:14:51a a and are so we were able to
0:14:54provide the
0:14:55perceptual the position that that Q of the beginning of all of the beginning of this two
0:15:01uh a because we have use a a a a a that T is that the core do you know
0:15:06that was emission that that being or maybe proposed by a set in the last
0:15:10i i then basically we were able to
0:15:12perception of like the propositions that for each he's
0:15:15for each frequency could we don't
0:15:17and the positions
0:15:19decoder
0:15:20i
0:15:21uh is actually
0:15:23uh occur
0:15:25the quantization step
0:15:28has this advantage will show increasing computational complexity
0:15:32and uh show good plane to introduce source of the temporal masking mechanisms
0:15:37in the J email
0:15:39a a subjective assessment campaign or to the data with
0:15:43we have obtained a
0:15:45a double check metrics and uh
0:15:47but also performing a perceptual optimization of the that
0:15:50for that
0:15:51a be in the coding
0:15:53so these ends my presentation of what i think you for
0:15:56actually
0:15:57yeah
0:16:00so we have
0:16:01i to question
0:16:07you use of the ball
0:16:08some be used
0:16:09a G C
0:16:11actually we start that what it is not that they uh
0:16:16the end of the call for all
0:16:18but uh uh the prayer and i think that the problem is that they have to
0:16:23a a form another subject to campaign which is
0:16:26but time so as you know two
0:16:29find that the the call one
0:16:31actually
0:16:32actually i have a of G
0:16:34looks so because it it this point so uh a
0:16:37statistic indicate the with east which is called metric the solving our but
0:16:42i
0:16:42we have a plane
0:16:52so
0:16:55so
0:16:55spatial uh
0:16:58this
0:16:59you use already available to
0:17:02or you want to
0:17:03since simple
0:17:05the
0:17:06for that will you
0:17:08from subject
0:17:09hmmm
0:17:11a a uh do so the the the and
0:17:15special temporal on sounds
0:17:17it in asking which is based on the the little T
0:17:21all the the has some there as which are that and so on and and the high and money for
0:17:27that yeah
0:17:28independent from the each E C
0:17:30and the also uh
0:17:34well
0:17:35you need to the motion but
0:17:37for you don't need to extend to
0:17:40cheese
0:17:41coding tools you can
0:17:42use the model
0:17:44T
0:17:44with
0:17:46performing i
0:17:47actions
0:17:48because the that
0:17:50the time the the
0:17:51if you see
0:17:53we go back to use
0:17:54maybe
0:18:04go back to this equation
0:18:07but you close to the the is is the special thing
0:18:10in a mean but it to a for them more that we have used
0:18:13he's the proposed by professor non doing
0:18:16the called university
0:18:17you just multiply to the J and temporal to a
0:18:20and that to
0:18:22uh
0:18:24scott as i remember that not
0:18:25imply a an extension to worse
0:18:28the the coding to
0:18:29each
0:18:34and
0:18:34okay
0:18:35thank Q