you your

in if

i

that

i

i

you

a most

i

i

makes use at

operates

because the to that

it's techniques

don't incorporate

can

in a and he's he's

a

or there techniques include the neighbourhood thing and that

but all of these techniques

and

well

in the present so so those or and the

a a a a set nation change

i

all this model

i has to be at that the i don't time

because the scene of all

you know where to start the you we propose a every young and the the technique

which include the context in the analyses

and also we proposed a way to get base frame what

it's such a way that we have a

not just colour information through the approximation coefficients

but also

it a a for the data

information

we propose a um what the resolution framework

so we have a information at different or solution they've

a a

all where part um model has also to be a it to the one percent of the scene

the first that's that in a where approach to is and matched segmentation

we are not used to now we are not willing to is the in the is a segmentation is that

the T E use

we had just think that is taking it suit

a a a a a what is start

a to use a a a a a a the quality of their so so the problem you my

what it to come in in mass we but fun a a we let this composition

a a a a a a three level wavelet there's composition

a a you a week a a a as a a a a set of approximation coefficients for a a

T M B band

and also a a set of detail coefficients a a for the great matt's in

what you on that of where to cut and the are one

we select the a a uh are to one now and wavelet that we a symmetric response

and then a support with of nine peaks

a after the way that uh there's a composition of forty two incoming brady and we have a a set

of approximation and detail coefficients

a

we model the distribution of the coefficients using a parametric model

in have to a more that in we have a a future vector

we the parameters of the model

so what their estimation coefficients

we select a are R R C and you would see "'em" why

because a a the approximation coefficients of data was also some of it

i so

boast to have a uniform all

in this case that's a feature vector we have a

the mean and its parents

for a air the M B band

in the case of the coffee coefficients

we select that a a a a a based on the work of my that

i can add lies gaussian distribution

in this case for it's set up

oh yeah and detail coefficients we caff the ad

self self i'm be that are to me

yeah

you can see an example of the i where a young cat of the scheme

a

this is a select like re on at X to region

in the second call and you can see

a a the in blue

they these two will shown of the approximation coefficients in the air G and B back

and and right that that pretty size these motion

in the third column

yeah you can see the a a a a a and this two was gonna be oh vertical the kind

of detail coefficients

for the level and one two at three

and in red

a in in blue

the a

the the to decide

as to motion

as you kind of sir

the a

they them more than

follows the a the messy of the real stiff was

a

in in order to me certain them a similarity between two radio

between a a a a a is the incoming and and the predicted re

we need a a a a at this time

no what case we have selected a will back like their time

a a which provide a a it because solution

for a a a a a a a a approximation off a coefficients

uh or or absent is to was john for to a

yeah can analyse sound as to which on what data type course

well at there a at a nice and of the radio

yeah i i now we have a a a a are more than

we have selected a a a a a uh and the strategy for

bottom of the based on this that you proposed by a still for at all

which operate at picks so that

now what case a a where i i don't a on is small that it

like i you start of a M all

in where it's small

a a more this

the D V A C on of the of each instantiation of the way you

with

spec

to a

to the uh a a pretty the rate them so in these

question

D

is that these stands school day of fading of a instant they should not be a in based in that

they that

with this fact

to they pretty of the whole

of the mixed

a

or mean that

is that weight being or i factor that indicates to really have S all these model it cost of the

car

in the race and you story of a region

we have cute

the distance

like how we in some between approximation the these you of the approximation coefficients and that is that's between to

detect which scenes

a

based on a yeah that it do that if is yes that and it is sensitive to sudden illumination

chain

so

a a a a eighteen and he match

we compute

you this time

two

with

all of the cake um all of the a

or a mixture

E

these distance he's

and data up score level value

is rate you on these incoming common break you maps with the are well

then we have to update all the twenty five dollars

it when

a a a a a

all the learning cost that

a fast and eight yeah

of of the a a does it the of a that they still

oh of a a a a um

more

to change

and then we have that they of they but by me there's of the all that

of the a at model

in this case we use another done

a we use proportion not don't lead to the learning was done but also

to a probability

or or a instant annotation of the gradient and of be longing

to the a a

as we open rate that the region they of it one only one of the modes is going to be

representative of the power

it is small

ah

the a

hi as the way you can cost than because she's the more relevant in that in and story of the

real

and also is the model stuff

because it's the standard deviation is still

but when the you mode that that are present

a i changes

the a suppose

a a that uh

configurations in high

had a happens

so we need to make

are segmentation

now we are going to present some result

oh or whatever i

and the a a determination change is and the sudden illumination change keys

in the presence of such a a and when and not come part

a

okay so that but it a nation change C

a

this C in the at based people or a profile of the change we have analyzed

a the low

oh a change

is

slow were compared with the velocity of the

adaptation of their background model

in the first plot you can see in red

the distance tendencies

well each in rate don't to a predict the brady on i don't time

a

this distance is approximately constant

and know

has a a a a a a low a you

that for so on a known C point

and this is because of the use

of the use of they could by like that is that we

which is not a are are a a we be

so we speak a or is not to linear

and in this

in the second that you can see that way the associated

to the models

of the gaussian of the mixture of colours and but we only have one model

so it's waiting in fact to reduce so was one

in the third block you can see the evolution of the standard deviation

a associated to the principal and model

a

in the initialization of press that's we start with that high

a standard deviation bay you

but

a a a a a a a when the a

on there there N is

a is this just table

yeah the standard deviation of the cries

and T to reach is

a

this as a duration the

asr conclusion a them um they're model source the change

and the parameters that describe the send that they've more at T to perfect

a

in this case we are going to compare the um

okay

this is E G B space corner select it with the a you be space or

a and

in

that distance that's i don't i in and that the gradual elimination change

is kind are in the

and uh

with respect to

to the you can be but so we can think that the a a a a a a you be

expected to

are you be colour space

in and the rather than an image in the change

this is true but

it we can see the same week at you we yeah and that unknown illumination change

so uh in which the other the he's

a

the these that's in that file inside her and that is task in the U R E but

so we conclude that

we used the in need that it side information because a hmmm hmmm before making any decision

now we are going to start the response of a where approximation

when there is a sudden illumination change

we are going to use that the the influence of the a come map by to me to

if we call

if we keep the the same data asked to approximation and detail coefficients we to this

do results

in this court

we see that when this sudden illumination change that happens

the distance increases

and

and you model in green

a errors in the mixture of yeah

these new model it's

get get more than

and it's to standard deviation of the guys so fine it we we are a to detect uh

a set in the computation of the scene

but if we keep

more read advanced to the type with P C and

that is that is when this sudden in um change happens

is a and that the S one level so

not model

a in the a in the mixture C

so we can want to do

that a

a a a a a a little are where

in the the that of how what or uh of our what are you in

is dependent on the pay you um a

a a we we should is still a low know what a low weight use of them

and

sudden illumination change

what peaks based on what then

i i it to and that with the presence of seven

a you know what are approach the inclusion of the on this and the that information

i low so as to and with this press

uh

so on

we

B as

in these court

that is done is approximately constant in the presence of subtle

we have only one principal more in the mixture

and the

is is the standard deviation or on

maybe a your is

a

i

are approximately constant

find are going to present some results

in a when uh

a when there is a a and you all get in the scene that just incorporate it in the back

in this case when the change happens

a a that these stands

a is

in in in in not in no approximation coefficient but

especially

in the detail coefficient

so

and new model of years

in in in green in these case

and a

so and

at the beginning there is standard deviation none of these new small

he's kind

but

when the C a C on

all this scene is

a table

this standard deviation the guys

and

and and you know a a a a a change in the print about all

how and so now

we decide that they that is a new configuration of their scene and we will uh

the will just a a a a a the segmentation of the re

now we are going to show you the behaviour of a word or problem

is win new well like it's a years in the scene

i

uh

new that's set B are where

or a a uh peers in the scene

in the scene

they are also at a rate

now

but they can be compared

the power

the try and they just plan a sort or now so that they do not affect the principle of all

new people are with in uh in the scene

with C

that

i

but so where the region

not affect in also the a principal model

when

a a it to work was cross over there rate and the behavior now it's the same

and the that the previous one

and for N

we are going to so you

the performance of their but thing when and new you did a a in this thing um become calm

uh

the person on and there's in this scene

and remains

there are on time and new model appear in the mixture of corruption

the it becomes

a a more significant

a

and now

that the the principal

more

so to compute the joint consideration of approximation and detail coefficients to model the we use and mixed of was

send set low

not only handling efficiency we see illumination change

i also with a a with new all that is that the as in the scene um become part of

the background

uh a with the present of so

the information got are in the proposed framework so how but that is there not only for problem of more

than

but also for intelligent analysis of the scene lucien

so

thank you

so i to at all

i think

what we shall four

that was able to to each

i

how you all

and i mean

all channels you a little and you don't watch a little

no quarter training

which about

but

for

research

a

this

so

i

from

oh

size

it was

a small

oh

or or or or

i

i

so

no but

we have to take the that that the real nice

be

i

where you need more at noise was are more

a a a a a a P A O T it coming to the reagan

to be detected

so we would have to read using in these case the say of the you

you you need or something like that

which

stop

i

a

right

so as use uh we uh

like for or multiple also was wonderful

the was so

for for me

a a a a or a question

three calls is such we sure

i