0:00:13you your
0:00:17in if
0:00:27a most
0:00:29makes use at
0:00:31because the to that
0:00:33it's techniques
0:00:34don't incorporate
0:00:36in a and he's he's
0:00:39or there techniques include the neighbourhood thing and that
0:00:43but all of these techniques
0:00:46in the present so so those or and the
0:00:49a a a a set nation change
0:00:54all this model
0:00:56i has to be at that the i don't time
0:00:59because the scene of all
0:01:04you know where to start the you we propose a every young and the the technique
0:01:08which include the context in the analyses
0:01:11and also we proposed a way to get base frame what
0:01:14it's such a way that we have a
0:01:16not just colour information through the approximation coefficients
0:01:20but also
0:01:21it a a for the data
0:01:25we propose a um what the resolution framework
0:01:28so we have a information at different or solution they've
0:01:32a a
0:01:34all where part um model has also to be a it to the one percent of the scene
0:01:41the first that's that in a where approach to is and matched segmentation
0:01:46we are not used to now we are not willing to is the in the is a segmentation is that
0:01:52the T E use
0:01:53we had just think that is taking it suit
0:01:56a a a a a what is start
0:01:58a to use a a a a a a the quality of their so so the problem you my
0:02:07what it to come in in mass we but fun a a we let this composition
0:02:12a a a a a a three level wavelet there's composition
0:02:15a a you a week a a a as a a a a set of approximation coefficients for a a
0:02:21T M B band
0:02:23and also a a set of detail coefficients a a for the great matt's in
0:02:27what you on that of where to cut and the are one
0:02:30we select the a a uh are to one now and wavelet that we a symmetric response
0:02:36and then a support with of nine peaks
0:02:41a after the way that uh there's a composition of forty two incoming brady and we have a a set
0:02:46of approximation and detail coefficients
0:02:50we model the distribution of the coefficients using a parametric model
0:02:56in have to a more that in we have a a future vector
0:03:00we the parameters of the model
0:03:03so what their estimation coefficients
0:03:06we select a are R R C and you would see "'em" why
0:03:09because a a the approximation coefficients of data was also some of it
0:03:14i so
0:03:15boast to have a uniform all
0:03:18in this case that's a feature vector we have a
0:03:20the mean and its parents
0:03:22for a air the M B band
0:03:25in the case of the coffee coefficients
0:03:27we select that a a a a a based on the work of my that
0:03:31i can add lies gaussian distribution
0:03:34in this case for it's set up
0:03:36oh yeah and detail coefficients we caff the ad
0:03:40self self i'm be that are to me
0:03:45you can see an example of the i where a young cat of the scheme
0:03:51this is a select like re on at X to region
0:03:56in the second call and you can see
0:03:58a a the in blue
0:04:01they these two will shown of the approximation coefficients in the air G and B back
0:04:06and and right that that pretty size these motion
0:04:09in the third column
0:04:11yeah you can see the a a a a a and this two was gonna be oh vertical the kind
0:04:16of detail coefficients
0:04:18for the level and one two at three
0:04:21and in red
0:04:22a in in blue
0:04:23the a
0:04:24the the to decide
0:04:26as to motion
0:04:28as you kind of sir
0:04:29the a
0:04:31they them more than
0:04:32follows the a the messy of the real stiff was
0:04:41in in order to me certain them a similarity between two radio
0:04:45between a a a a a is the incoming and and the predicted re
0:04:50we need a a a a at this time
0:04:52no what case we have selected a will back like their time
0:04:56a a which provide a a it because solution
0:04:59for a a a a a a a a approximation off a coefficients
0:05:04uh or or absent is to was john for to a
0:05:07yeah can analyse sound as to which on what data type course
0:05:14well at there a at a nice and of the radio
0:05:17yeah i i now we have a a a a are more than
0:05:21we have selected a a a a a uh and the strategy for
0:05:24bottom of the based on this that you proposed by a still for at all
0:05:28which operate at picks so that
0:05:32now what case a a where i i don't a on is small that it
0:05:36like i you start of a M all
0:05:39in where it's small
0:05:41a a more this
0:05:42the D V A C on of the of each instantiation of the way you
0:05:48to a
0:05:49to the uh a a pretty the rate them so in these
0:05:55is that these stands school day of fading of a instant they should not be a in based in that
0:06:01they that
0:06:02with this fact
0:06:03to they pretty of the whole
0:06:06of the mixed
0:06:08or mean that
0:06:09is that weight being or i factor that indicates to really have S all these model it cost of the
0:06:16in the race and you story of a region
0:06:20we have cute
0:06:21the distance
0:06:22like how we in some between approximation the these you of the approximation coefficients and that is that's between to
0:06:29detect which scenes
0:06:32based on a yeah that it do that if is yes that and it is sensitive to sudden illumination
0:06:40a a a a eighteen and he match
0:06:42we compute
0:06:43you this time
0:06:46all of the cake um all of the a
0:06:49or a mixture
0:06:51these distance he's
0:06:53and data up score level value
0:06:56is rate you on these incoming common break you maps with the are well
0:07:01then we have to update all the twenty five dollars
0:07:05it when
0:07:06a a a a a
0:07:08all the learning cost that
0:07:10a fast and eight yeah
0:07:11of of the a a does it the of a that they still
0:07:14oh of a a a a um
0:07:17to change
0:07:19and then we have that they of they but by me there's of the all that
0:07:22of the a at model
0:07:24in this case we use another done
0:07:27a we use proportion not don't lead to the learning was done but also
0:07:32to a probability
0:07:35or or a instant annotation of the gradient and of be longing
0:07:39to the a a
0:07:45as we open rate that the region they of it one only one of the modes is going to be
0:07:49representative of the power
0:07:51it is small
0:07:53the a
0:07:54hi as the way you can cost than because she's the more relevant in that in and story of the
0:08:01and also is the model stuff
0:08:03because it's the standard deviation is still
0:08:07but when the you mode that that are present
0:08:10a i changes
0:08:11the a suppose
0:08:14a a that uh
0:08:16configurations in high
0:08:18had a happens
0:08:19so we need to make
0:08:21are segmentation
0:08:26now we are going to present some result
0:08:29oh or whatever i
0:08:30and the a a determination change is and the sudden illumination change keys
0:08:36in the presence of such a a and when and not come part
0:08:42okay so that but it a nation change C
0:08:46this C in the at based people or a profile of the change we have analyzed
0:08:51a the low
0:08:52oh a change
0:08:55slow were compared with the velocity of the
0:08:58adaptation of their background model
0:09:01in the first plot you can see in red
0:09:05the distance tendencies
0:09:06well each in rate don't to a predict the brady on i don't time
0:09:12this distance is approximately constant
0:09:15and know
0:09:17has a a a a a a low a you
0:09:19that for so on a known C point
0:09:24and this is because of the use
0:09:26of the use of they could by like that is that we
0:09:29which is not a are are a a we be
0:09:32so we speak a or is not to linear
0:09:35and in this
0:09:37in the second that you can see that way the associated
0:09:40to the models
0:09:41of the gaussian of the mixture of colours and but we only have one model
0:09:46so it's waiting in fact to reduce so was one
0:09:50in the third block you can see the evolution of the standard deviation
0:09:54a associated to the principal and model
0:10:00in the initialization of press that's we start with that high
0:10:04a standard deviation bay you
0:10:07a a a a a a a when the a
0:10:09on there there N is
0:10:11a is this just table
0:10:13yeah the standard deviation of the cries
0:10:15and T to reach is
0:10:18this as a duration the
0:10:21asr conclusion a them um they're model source the change
0:10:26and the parameters that describe the send that they've more at T to perfect
0:10:34in this case we are going to compare the um
0:10:39this is E G B space corner select it with the a you be space or
0:10:46a and
0:10:49that distance that's i don't i in and that the gradual elimination change
0:10:53is kind are in the
0:10:55and uh
0:10:56with respect to
0:10:57to the you can be but so we can think that the a a a a a a you be
0:11:02expected to
0:11:03are you be colour space
0:11:05in and the rather than an image in the change
0:11:09this is true but
0:11:11it we can see the same week at you we yeah and that unknown illumination change
0:11:16so uh in which the other the he's
0:11:20the these that's in that file inside her and that is task in the U R E but
0:11:25so we conclude that
0:11:27we used the in need that it side information because a hmmm hmmm before making any decision
0:11:35now we are going to start the response of a where approximation
0:11:39when there is a sudden illumination change
0:11:42we are going to use that the the influence of the a come map by to me to
0:11:46if we call
0:11:47if we keep the the same data asked to approximation and detail coefficients we to this
0:11:54do results
0:11:55in this court
0:11:56we see that when this sudden illumination change that happens
0:12:00the distance increases
0:12:03and you model in green
0:12:05a errors in the mixture of yeah
0:12:08these new model it's
0:12:10get get more than
0:12:12and it's to standard deviation of the guys so fine it we we are a to detect uh
0:12:17a set in the computation of the scene
0:12:21but if we keep
0:12:23more read advanced to the type with P C and
0:12:27that is that is when this sudden in um change happens
0:12:31is a and that the S one level so
0:12:34not model
0:12:35a in the a in the mixture C
0:12:39so we can want to do
0:12:40that a
0:12:43a a a a a a little are where
0:12:45in the the that of how what or uh of our what are you in
0:12:49is dependent on the pay you um a
0:12:52a a we we should is still a low know what a low weight use of them
0:12:58sudden illumination change
0:13:03what peaks based on what then
0:13:05i i it to and that with the presence of seven
0:13:10a you know what are approach the inclusion of the on this and the that information
0:13:15i low so as to and with this press
0:13:18so on
0:13:21B as
0:13:23in these court
0:13:24that is done is approximately constant in the presence of subtle
0:13:29we have only one principal more in the mixture
0:13:33and the
0:13:35is is the standard deviation or on
0:13:37maybe a your is
0:13:41are approximately constant
0:13:45find are going to present some results
0:13:48in a when uh
0:13:49a when there is a a and you all get in the scene that just incorporate it in the back
0:13:55in this case when the change happens
0:13:58a a that these stands
0:14:00a is
0:14:02in in in in not in no approximation coefficient but
0:14:07in the detail coefficient
0:14:10and new model of years
0:14:13in in in green in these case
0:14:16and a
0:14:17so and
0:14:18at the beginning there is standard deviation none of these new small
0:14:23he's kind
0:14:25when the C a C on
0:14:27all this scene is
0:14:29a table
0:14:30this standard deviation the guys
0:14:33and and you know a a a a a change in the print about all
0:14:36how and so now
0:14:38we decide that they that is a new configuration of their scene and we will uh
0:14:44the will just a a a a a the segmentation of the re
0:14:51now we are going to show you the behaviour of a word or problem
0:14:56is win new well like it's a years in the scene
0:15:17new that's set B are where
0:15:20or a a uh peers in the scene
0:15:23in the scene
0:15:25they are also at a rate
0:15:30but they can be compared
0:15:31the power
0:15:36the try and they just plan a sort or now so that they do not affect the principle of all
0:15:45new people are with in uh in the scene
0:15:47with C
0:15:58but so where the region
0:16:00not affect in also the a principal model
0:16:07a a it to work was cross over there rate and the behavior now it's the same
0:16:12and the that the previous one
0:16:21and for N
0:16:23we are going to so you
0:16:28the performance of their but thing when and new you did a a in this thing um become calm
0:16:36the person on and there's in this scene
0:16:39and remains
0:16:43there are on time and new model appear in the mixture of corruption
0:16:51the it becomes
0:16:52a a more significant
0:16:56and now
0:16:57that the the principal
0:17:06so to compute the joint consideration of approximation and detail coefficients to model the we use and mixed of was
0:17:12send set low
0:17:14not only handling efficiency we see illumination change
0:17:18i also with a a with new all that is that the as in the scene um become part of
0:17:23the background
0:17:24uh a with the present of so
0:17:27the information got are in the proposed framework so how but that is there not only for problem of more
0:17:32but also for intelligent analysis of the scene lucien
0:17:38thank you
0:17:44so i to at all
0:17:45i think
0:17:51what we shall four
0:17:54that was able to to each
0:17:58how you all
0:18:02and i mean
0:18:03all channels you a little and you don't watch a little
0:18:08no quarter training
0:18:22which about
0:18:45it was
0:18:47a small
0:18:50or or or or
0:18:55no but
0:18:57we have to take the that that the real nice
0:19:02where you need more at noise was are more
0:19:06a a a a a a P A O T it coming to the reagan
0:19:10to be detected
0:19:11so we would have to read using in these case the say of the you
0:19:15you you need or something like that
0:19:33so as use uh we uh
0:19:36like for or multiple also was wonderful
0:19:39the was so
0:19:41for for me
0:19:43a a a a or a question
0:19:46three calls is such we sure