| 0:00:13 | a |
|---|
| 0:00:14 | the and they have an uh i'm and |
|---|
| 0:00:16 | a uh i'm from the that's going from a the problem |
|---|
| 0:00:19 | uh and you estimated quite spar |
|---|
| 0:00:22 | and that type on map and bases |
|---|
| 0:00:24 | is a reason against sport image you the creation |
|---|
| 0:00:27 | and this voice collaborate with uh |
|---|
| 0:00:30 | yeah and got and the for a radio |
|---|
| 0:00:33 | so uh |
|---|
| 0:00:34 | yes the a around this print is and uh for our uh uh uh introduced the brief can set up |
|---|
| 0:00:39 | a you listen again |
|---|
| 0:00:41 | and then and i have a |
|---|
| 0:00:42 | at discussed how to formulate the image into to and you is again |
|---|
| 0:00:47 | and by are so some in is spends a read all for or by some conclusions |
|---|
| 0:00:53 | so uh seventy again |
|---|
| 0:00:55 | maybe you C components |
|---|
| 0:00:57 | and the first compress prayer who prayed again |
|---|
| 0:01:01 | and here is by note the base you one up to you and we assume they are and prayers S |
|---|
| 0:01:06 | and this and compare use the use straight is set |
|---|
| 0:01:09 | and and here |
|---|
| 0:01:10 | a a a i ten for the few straight is set for use at you Y |
|---|
| 0:01:15 | and the |
|---|
| 0:01:16 | that that i uh i think so as a at a i i is the |
|---|
| 0:01:21 | all of possible uh a Q strategy that that use so you like can pay |
|---|
| 0:01:27 | and this can put a is the feel function |
|---|
| 0:01:30 | we is uh with determine the pale that that use so you uh a that use the can opt ten |
|---|
| 0:01:35 | by taking a something and |
|---|
| 0:01:37 | a so or the add i stand for the uh |
|---|
| 0:01:41 | little pale for a single of you uh you i |
|---|
| 0:01:45 | and with a note is that uh all the at i |
|---|
| 0:01:48 | user is not only to turn you by le use you ice X |
|---|
| 0:01:52 | but also all allows users X |
|---|
| 0:01:55 | and and yeah i us some a to do this and that we use in this print |
|---|
| 0:01:59 | first is the strategy profile |
|---|
| 0:02:02 | a |
|---|
| 0:02:03 | is that class less and all all the players X and you want to a a and |
|---|
| 0:02:07 | and this strategies is it a file is set pretty a |
|---|
| 0:02:11 | a you i use a neck i |
|---|
| 0:02:13 | uh so is the and it again the class and of the all operators |
|---|
| 0:02:17 | uh actually it's seven player you wise X |
|---|
| 0:02:21 | and the P say set as i say is that content all of possible pure strategy let the use that |
|---|
| 0:02:27 | you i can take |
|---|
| 0:02:28 | and that means strategy G is a random might a strategy |
|---|
| 0:02:32 | uh well |
|---|
| 0:02:33 | can be |
|---|
| 0:02:34 | represents as pi |
|---|
| 0:02:36 | we is a probability distribution of these |
|---|
| 0:02:39 | Q trace set |
|---|
| 0:02:42 | so uh |
|---|
| 0:02:44 | to mess my on the apple and pale a prayer and it's know all information all are out of videos |
|---|
| 0:02:51 | uh and channel is very difficult or even impossible to for prayer that's and no out a real a strategy |
|---|
| 0:02:57 | and P |
|---|
| 0:02:59 | so i that's K |
|---|
| 0:03:00 | to improve the P L |
|---|
| 0:03:02 | a a a a uh we are let hard to predict the N using the um um target to you |
|---|
| 0:03:07 | uh |
|---|
| 0:03:07 | oh understanding by then uh by building |
|---|
| 0:03:11 | and this this to the concept you loose in again |
|---|
| 0:03:14 | so and people is again is that and that's not be the interest and dependent strategy in population |
|---|
| 0:03:21 | uh and uh a was set in that you recent again is slightly bruce and they stable strategy |
|---|
| 0:03:26 | which is |
|---|
| 0:03:27 | but strategy is that's that if all band bounds of the part this um top it |
|---|
| 0:03:32 | then no me intense that you can invade the populations and that in inference |
|---|
| 0:03:37 | well mistress and |
|---|
| 0:03:39 | and such and |
|---|
| 0:03:40 | and you produce an is they was to you can be simply and using the updating function here |
|---|
| 0:03:46 | and |
|---|
| 0:03:46 | uh the idea uh behind is updating dating thing is that |
|---|
| 0:03:50 | uh the probability of using us that ten |
|---|
| 0:03:53 | strategy |
|---|
| 0:03:54 | at time T plus one |
|---|
| 0:03:56 | is it to me is proposed a note to look pay all |
|---|
| 0:04:00 | oh using using this strategy a previous time |
|---|
| 0:04:04 | T |
|---|
| 0:04:06 | so a with the background of the you risen again now you're like uh should discuss how to uh for |
|---|
| 0:04:12 | me the image into and as a duration again |
|---|
| 0:04:15 | so i think you in their present problem we have the |
|---|
| 0:04:19 | uh a less as a image uh i here the |
|---|
| 0:04:23 | but so but i know the low less was and not that is |
|---|
| 0:04:26 | two |
|---|
| 0:04:27 | can a high does to the image |
|---|
| 0:04:30 | and and in it is a rate are they ask |
|---|
| 0:04:32 | many but um is that mean proposed for this problem |
|---|
| 0:04:35 | and here we only consider the uh was best so |
|---|
| 0:04:38 | but that the if you're attending |
|---|
| 0:04:40 | well for it's P so i had a it's uh for it uh it's be sides i see |
|---|
| 0:04:46 | uh we use uh |
|---|
| 0:04:47 | a in a combination of a set of uh a low less and |
|---|
| 0:04:51 | this so i out of here |
|---|
| 0:04:54 | i use is we assume L uh ways and non negative and this sum up to one |
|---|
| 0:04:59 | and and if we know the this that a the low less and P so yeah the problem not be |
|---|
| 0:05:05 | how to find a way |
|---|
| 0:05:07 | so uh or is that we know that the team like can not be of then |
|---|
| 0:05:11 | because um |
|---|
| 0:05:13 | we do not have a or in a high uh less to see me image |
|---|
| 0:05:16 | uh and a look up previous approach |
|---|
| 0:05:20 | i i estimate the weights |
|---|
| 0:05:21 | based on the sounds and that the |
|---|
| 0:05:23 | could balance makes use of a high does a |
|---|
| 0:05:26 | you means it can be where estimate form no combat about its you of low low say image |
|---|
| 0:05:31 | i one is the sounds a not true then the in to present a performance |
|---|
| 0:05:36 | we have be great thirty degraded |
|---|
| 0:05:38 | so he a |
|---|
| 0:05:40 | is that a a in the way in one that |
|---|
| 0:05:42 | we propose to progress the we find the way by at an antibody that the made in the ways based |
|---|
| 0:05:49 | on consistent |
|---|
| 0:05:50 | and then finding the look and and using the ways |
|---|
| 0:05:53 | as as a refinement process as is at sri and you produce in a process |
|---|
| 0:05:57 | so which can be naturally really one it S and you resume again |
|---|
| 0:06:02 | so uh |
|---|
| 0:06:03 | yeah is the uh units there for me i say it's very simple |
|---|
| 0:06:07 | we you see that a a in this uh a one is and the prayer a to a no high |
|---|
| 0:06:11 | does of them so |
|---|
| 0:06:12 | i it's i G |
|---|
| 0:06:14 | and the use i T are the corresponding little as an piece so i have here are |
|---|
| 0:06:19 | and that use as that is that can do this ad |
|---|
| 0:06:23 | and that me is that the uh the outcome of that means strategy is the estimate of the high does |
|---|
| 0:06:28 | and piece shows i had it's |
|---|
| 0:06:32 | and the probability in the this strategy can be do guys the negative |
|---|
| 0:06:37 | uh norm rice a topic you |
|---|
| 0:06:40 | so we so kind the formulation |
|---|
| 0:06:42 | uh the problem of in the a |
|---|
| 0:06:45 | S the mail to find a an piece so we can the problem of finding the people produce a stable |
|---|
| 0:06:49 | strategy of this again |
|---|
| 0:06:53 | so a a how the size now look Q said to hear a uh we use use a very simple |
|---|
| 0:06:58 | up used strategy said |
|---|
| 0:06:59 | well for it's a a high to and be so we just use the the |
|---|
| 0:07:03 | uh |
|---|
| 0:07:04 | a square window of by than the is low less system this shows |
|---|
| 0:07:08 | example here uh uh we have the |
|---|
| 0:07:11 | i sh uh this uh this we uh a hard as and P is uh i |
|---|
| 0:07:15 | two and my as one to and |
|---|
| 0:07:17 | and i two and two and minus one and i two and to N |
|---|
| 0:07:22 | and with just use the scroll window as and about a |
|---|
| 0:07:25 | the P so look case that's a as one and that's one minus one |
|---|
| 0:07:31 | and |
|---|
| 0:07:32 | for a pale constant and a |
|---|
| 0:07:35 | we we force uh define the |
|---|
| 0:07:39 | similarity measure between uh can it is uh |
|---|
| 0:07:43 | twenty it P so the i C and the time P so i it's i G |
|---|
| 0:07:47 | and these uh uh uh |
|---|
| 0:07:49 | similar he's so that's by lot you could eat it this and of the |
|---|
| 0:07:54 | oh local pet sent the by the i'd G |
|---|
| 0:07:56 | which is be yeah i G |
|---|
| 0:07:59 | and that of the local patch since the by i it's i'd say we is to be i side G |
|---|
| 0:08:05 | and from this uh a a a it's of the and we can see that if a i'd T close |
|---|
| 0:08:09 | to the i as i'd say then that source and we're be |
|---|
| 0:08:12 | um a small |
|---|
| 0:08:14 | is that's K with to give a higher payoff O |
|---|
| 0:08:17 | a a high P L to that uh can it be so |
|---|
| 0:08:20 | and only out a hand if a i'd say is far away from i i say |
|---|
| 0:08:25 | then this just all since to be lots |
|---|
| 0:08:27 | and use that's K we give a low pay to the |
|---|
| 0:08:31 | look can the piece source |
|---|
| 0:08:33 | so is all sets uh in you uh in deviation a lapel since to be a six creasing function of |
|---|
| 0:08:38 | this D yeah G and are we use the as one in seconds constant |
|---|
| 0:08:44 | and |
|---|
| 0:08:45 | was a know this so that for uh |
|---|
| 0:08:47 | can it is to look at a hard put used an uh which actually use the we and just a |
|---|
| 0:08:53 | be at times no uh time and S |
|---|
| 0:08:55 | i had a it's T I G |
|---|
| 0:08:58 | and if we compute a you could it this than you would be zero |
|---|
| 0:09:01 | but that's to the beacons that's some may have some that also |
|---|
| 0:09:05 | we cannot use that um as a matter |
|---|
| 0:09:08 | so he we we use uh a very simple is that |
|---|
| 0:09:11 | just we is that this also using the minimal distortion or or or a duck is so |
|---|
| 0:09:17 | and again let P L forms an all is used to compute by a a once and |
|---|
| 0:09:24 | so uh |
|---|
| 0:09:27 | so that's look at some uh i spend the out so we test all mess so |
|---|
| 0:09:31 | a to you need to did not board |
|---|
| 0:09:33 | and we assume that lower low those an image is is down sampled from a high does and you image |
|---|
| 0:09:38 | and we assume or is an image E |
|---|
| 0:09:42 | is a battle |
|---|
| 0:09:44 | for a a a as a been marked |
|---|
| 0:09:46 | and we compare with the by Q P and S A are misses |
|---|
| 0:09:51 | so if here is uh the result for a D a V G a |
|---|
| 0:09:55 | the A the low lows the image can be is the result of open by by Q P |
|---|
| 0:10:00 | and C is an a result uh generated by a the i and D is the |
|---|
| 0:10:05 | read all all misses |
|---|
| 0:10:07 | we can see that uh |
|---|
| 0:10:10 | uh B C uh the are some a big at fair here |
|---|
| 0:10:14 | uh we saw mess so we can a voice that's kinds are but |
|---|
| 0:10:17 | maybe can the good is more clear |
|---|
| 0:10:20 | uh for a boat image |
|---|
| 0:10:22 | we can see that they are some fig for that |
|---|
| 0:10:26 | we we a mass so we can uh a great the uh read you this kinds are the pen but |
|---|
| 0:10:30 | you'd better read |
|---|
| 0:10:32 | Y thing which T is that is um |
|---|
| 0:10:35 | in the original boat image yeah some since noise |
|---|
| 0:10:40 | and |
|---|
| 0:10:41 | in that case if uh you |
|---|
| 0:10:43 | use a a a i to be and it always yeah i to separate such kind of image the noise |
|---|
| 0:10:48 | would be puts i'll and this at to use re though closing the figure that out that |
|---|
| 0:10:54 | and |
|---|
| 0:10:55 | you we saw mess so we we can as we can suppress such kinds of sense noise |
|---|
| 0:10:59 | which being used as a a mess that have the potential to perform joint in to and |
|---|
| 0:11:05 | and and this is a body if i to using this is but to here |
|---|
| 0:11:09 | a a we assume uh |
|---|
| 0:11:11 | less less as say just noise |
|---|
| 0:11:13 | it's it's to buy some calls and noise |
|---|
| 0:11:16 | uh and we compare our our that we sit them as a lap was performed a denoising using the long |
|---|
| 0:11:22 | local and then performed in to place and using |
|---|
| 0:11:25 | uh by Q be all the |
|---|
| 0:11:27 | as |
|---|
| 0:11:28 | we can see that the S some are right |
|---|
| 0:11:30 | you |
|---|
| 0:11:31 | well the phase and the no it's region |
|---|
| 0:11:34 | and we |
|---|
| 0:11:34 | and this kind of a a a page uh maybe come um many from |
|---|
| 0:11:40 | uh this separation of that the anointing a to prison |
|---|
| 0:11:44 | uh |
|---|
| 0:11:45 | which means that you you you or did know any i wouldn't can not if we moves noise |
|---|
| 0:11:50 | then the noise would be pushed are you in in the in the present process |
|---|
| 0:11:54 | i we saw a mass so uh we can |
|---|
| 0:11:56 | performs time it ten is at the non you can to raise and and we such kinds of sewing a |
|---|
| 0:12:01 | process we can suppress such kinds of the play |
|---|
| 0:12:05 | so um |
|---|
| 0:12:07 | a come to can goes and uh so we propose a |
|---|
| 0:12:11 | but are from form actually a general framework or a image uh |
|---|
| 0:12:15 | in the present and using that you person again |
|---|
| 0:12:18 | and this kind of for what it's very simple |
|---|
| 0:12:20 | and and in a few that we like to investigate hall fall out uh that we can uh and uh |
|---|
| 0:12:26 | good P L problems and to improve the result because now a |
|---|
| 0:12:29 | the P L the as is very straight for just using the it read it |
|---|
| 0:12:34 | this than uh between the low local patch |
|---|
| 0:12:37 | and we also uh |
|---|
| 0:12:39 | uh we also in in |
|---|
| 0:12:42 | a a gem a to investigate the possibility of gen lighting but addition approach and all framework |
|---|
| 0:12:49 | and just can my |
|---|
| 0:12:51 | vol |
|---|
| 0:12:51 | the Q |
|---|
| 0:12:57 | right so |
|---|
| 0:12:58 | we we can't to thanks okay |
|---|
| 0:13:00 | ah |
|---|
| 0:13:01 | we can afford |
|---|
| 0:13:03 | quite a few number of questions |
|---|
| 0:13:09 | so |
|---|
| 0:13:10 | um and you go back to the land uh uh uh in its uh |
|---|
| 0:13:13 | yeah O |
|---|
| 0:13:14 | maybe a couple more so what's the T P M |
|---|
| 0:13:17 | psnr |
|---|
| 0:13:18 | i i i i so i i segments that this uh a eyes only for this three all i see |
|---|
| 0:13:22 | okay because i have never or or right before |
|---|
| 0:13:26 | uh |
|---|
| 0:13:26 | oh typically for the entire image are right five thirty six right right of L a little right okay |
|---|
| 0:13:32 | yeah he even for the right keep but as i for a whole image is no |
|---|
| 0:13:36 | it's key problems in that |
|---|
| 0:13:39 | he a very naive questions um |
|---|
| 0:13:42 | typically one we play games we compete |
|---|
| 0:13:45 | uh_huh and and the play was trying to compete right |
|---|
| 0:13:48 | uh |
|---|
| 0:13:49 | but he you are posing |
|---|
| 0:13:52 | the uh be interpolated pixels is play the right and i don't see why |
|---|
| 0:13:58 | okay okay uh as the competition so why to why needed so for the given to the are two kinds |
|---|
| 0:14:03 | of the many yeah why as long at which the uh pretty a compute which is a i'll a a |
|---|
| 0:14:09 | and one type of a not type is the culprit again |
|---|
| 0:14:12 | well i |
|---|
| 0:14:13 | pay of pretty which is a lower so |
|---|
| 0:14:15 | for for these will lose it again this the corporate again |
|---|
| 0:14:18 | so they call in step |
|---|
| 0:14:21 | compute you which is um |
|---|
| 0:14:23 | so yeah so |
|---|
| 0:14:25 | so basically what how they cooperate with these are is that uh |
|---|
| 0:14:30 | for example uh we use this uh |
|---|
| 0:14:34 | when we compute the |
|---|
| 0:14:35 | P of phones and |
|---|
| 0:14:37 | we were use a week can just and all |
|---|
| 0:14:39 | this just so for this so |
|---|
| 0:14:41 | a so in such case are if they have |
|---|
| 0:14:44 | uh |
|---|
| 0:14:45 | very high similarity then they we help each other to what you're better results |
|---|
| 0:14:50 | yeah |
|---|
| 0:14:53 | um |
|---|
| 0:14:54 | and |
|---|
| 0:14:55 | obviously is as |
|---|
| 0:14:57 | it changes |
|---|
| 0:14:58 | uh |
|---|
| 0:14:59 | from from one pixel of the uh the other the weighting the resulting optimal weights that you get out at |
|---|
| 0:15:05 | the end is |
|---|
| 0:15:07 | is |
|---|
| 0:15:07 | is uh as a function of each |
|---|
| 0:15:10 | pixel corps |
|---|
| 0:15:12 | not not by patch but by |
|---|
| 0:15:15 | yeah just for it's piece that the for each voice of right yeah wait yeah each pixel right |
|---|
| 0:15:20 | uh do they have |
|---|
| 0:15:21 | i all have to for little exactly the um |
|---|
| 0:15:25 | some up to one right you need to have some up to why how to wise you have |
|---|
| 0:15:30 | different being the intensity with a deeper |
|---|
| 0:15:33 | the was |
|---|
| 0:15:33 | each |
|---|
| 0:15:35 | any of the question |
|---|
| 0:15:38 | okay if not uh |
|---|
| 0:15:40 | hopefully a use right |
|---|
| 0:15:43 | okay |
|---|
| 0:15:44 | then |
|---|