yeah thank uh morning uh to they are like to send a paper a stretching an you're in this party compensated prediction and he's for multi view video coding so um he's the content or uh my presentation uh uh make a joint action of the disparity compensated in section especially that's the techniques used in the curve uh multi we the coding standard and you propose a lost yeah i forty for S E S for shot for the S yes six technique and the related uh consideration and some experimental sound would be given and the conclusion joint final so um P is the introduction for the current uh a multi view video coding a a standard that at to it is then that they exploit correlation between the real okay as a this fight previous uh H so um it is a a a a a a a extension of the explore as one a V C so uh it is is this uh in and so so the that this party uh information recall or that by the disparity compensated prediction because uh this technique station uh exposed to what you see they don't same coding instead all spot to the cool so um also so the the the coding to is just the same as they used in the and C P so um so uh if they use the same set of coding tools a they the only sport the pure translation uh um most a back to this party act so it is a broad based um um motion compensated prediction technique so that this party back to K send the the should all the blocks where uh but i not indicate the act uh a to that so um i'll met what is to put holes a a a technique to improve the this particle can see text for the and B C ah and the effect that um oh we want to cheap is to um buy by a a a a a new technique a proposed the course of a sub sample the block matching it does not imply a uh additional memory for storing the test additional we're frames and it you know need to perform a this to though uh and the polish so um all propose at any will be uh able to integrate to the actual to exploit is it so uh the S talk out the um a the this party fact uh first for four we propose that so a a uh i here we show a a a a a real world example or all a a a this party that it is a a real picture taken from the i a still can two can so it is obvious that um the the difference between the frames a are not just translation okay that's that uh to what you close to so for the uh oh example there is a wall uh uh on the on the right side and in many find that we find that uh for the last one but oh we use this one as the reference you can see that a on the wall the plate on the war it's up here to be a can press in the right view but so we find that a a there is a a compression in that if you a um not turning around you will be a stretching that so i here use the a vertical option for and and example uh for example in the C where some some horizontal objects on a if you look at two let me you can of fun this the texture on the see and you a that a a on the right real the the texture it uh appear to be uh a two a a i um um we said to have it is a a a a a a a year fact although though um you may not be a a a a um oh perfectly uh a processing uh more model that but we believe that um uh using the sheer in if can cross to me this that in this case so um so he he is the the uh uh uh S A six if that are we we want to a sprawl so it should be noted that a a a you the camera set this up and down is that the uh uh that right the the S S six if should be horse so yeah for should be a vertical instead or something so oh how do we provide a um a the S E X fig if fact and the uh uh this partly compensated prediction what in into what you i i that we can prove i a some uh uh are transformed the frame vol for the the the this part to text a a for example we can provide a some uh a she had that a a frame and also uh a strange uh a stretch and compress the frame and we can make use of these and frame to at all different uh uh i is sick fact to to improve the that the this on the text but ah uh this uh a what if that we we'll have a a a a a very serious uh but not for it major problem uh for for practical use because uh and this frame based on require a some um memory space uh a in the and code because you you need to provide a a a number all reference frame a a for you can you can for for the problem actually has one so um the number uh the then we we we'll be role a a a a a with the number all it's is actual uh can provide and also uh the the computing line and for providing this uh S A six frames is also a complex because you be two uh a a you the poll at of the reference frame that performing transformation for for the S is section five the frame based approach is not right so in this research we will uh a post match what uh to achieve the S A six this C P i i a new technology at a new to a uh and i call a a sample the block matching that's i set it to you know it does not require additional memory and additional look the so um uh i here is all on the S A six this if you a a in the extra two six for U V C are we have a a what a pixel uh a of us to use it and a in a typical uh echo that we will have a uh some four times in the for it to that and frame for for as to um form the problem matching and and uh the and the uh motion compensated prediction works so um in the in the and C P and this if you process um uh the echo the we sub same the the frame to taking the uh uh problem in the original scale because you you you in in with that in four times and subset wave by what times so that the scale change but you can a the location the a pixel location so we define a oh we first define this uh uh is a a a subsampling sampling rate five question okay it should be for simple um to achieve the the S A six uh effect we can modify this um a subsampling sampling re for um for for for different probably test to achieve the if that you like to sport so yeah use the um um um stress and compression um uh if X that we we decide so that we use okay we use the the a that's same we we define by this creation but that and you can see that and the um top left corner all of the the the matrix data um this is and compression a a problem we pacing the original for part to i in the corner so i we will achieve a a a strange and compression horizontal horizontal for example we do uh we we this up to do the that the S C problem at a with three we you have a the the subsampling grid shown in uh the left of figure you can see that um this some and one where you all the horizontal september we all the all i the reference frame three we go you can do this uh pixels for example if we we we are subject to you of the S C problem to i five right you have a is trash problem so here is the the stress and press so for this we we um uh you use this uh are also send bring re well by a S problem to so a by such doing by one or a we will have a a a a a a a right view that uh problem for to the row by the the you if at the left that that if that we does such a use some on on minus one i this is uh example of the search of course a to achieve a more a stretch and compression it she hearing um if we can be more uh meet a in the S C A S from so so there's a how we uh cheap that S E is six five the uh C P oh uh you can uh we like to all would post some my experimental result i here is the that a general condition that can pick a the the uh in the be people in the G M Z seventeen encoder so there a high profile we use some a video sequence that is quite a used a paul room X that what's a right now so uh a a the prime time thus does and for your information so for the um for the prime we can we use the for our uh and also we use uh X where we use H it and session okay let we the horse on those us a ten or in way from one to not and X as she ring levels will be used at but minus four to four i i should of that uh our for is currently only have high P well for for the uh also so that uh um and a is on the uh uh coding efficiency we we also sell to set the and one is um with all you frames uh i in this set um the let and that sequence the base will will be a i frame and the right real is all P frame the P frame only um a predicted by uh this party it and i set is a a hierarchical be frame with uh save and B frames that is by company in in uh a new techniques uh this is used to uh explore the effect all the all our proposed idea in some practical group so for the at first set all the uh um uh encoding we sell um we find that uh we have uh improvement in all case and uh uh F H B that's and it's all one point six nine send so uh this is yeah that this one point no so vol the the the coding structure we've a seven hierarchical B thing we we found that uh we still have improvement in in all this case but um but the improvement is uh uh a little bits a i it is possible that a the improvement i do by was beeping because we did up i uh no one B frame but the F H B and is you are out of the so you want to two two one to five then so ah a a he is close and this oh take so that we have uh a proposed post that in that one two all to that that the the S A S six back for the disparity compensated prediction we use the subset for the wrong match technique this is a um that's yes S brought i'll taking the from the interpolated frame so that we don the two two and and the memory and notation full uh taking in uh blocks and somehow T spot so a all for future work we find more um a and we we for both uh D C P A and C P for for or five to with the the proof the um formant a a a a a general video you to and you also find some possible that you know were from subsampled sample you oh this is one one is that they think the very much we maybe one or two which yeah why but that that is the most for things you already an estimation on as you a in our analysis is found that the p-frame frame use of the P right we and use only a uh in the in the and at the and we C and a moment the P frame is used to be uh predicted by the uh i i i I detection from the same view it is usually a out there are very uh well to so so the room for whole using the in the in the we detection may not be very very good and is not use that and i do you yeah use the don't frames of course not that i frame is in i intra prediction because you produce even though it terms of motion still do in did you all the intra frame we not use of that's the ah and i'm not that from but the standard itself but five a okay i guess it's a but i think is just a meeting a homography i as opposed to sharing yeah estimation actually is a a a and that what is try to provide some uh approximation all this uh a so called more to try formation by a very simple way so so that the if at all in the the transmission may not be uh uh or a to the uh uh uh generated but would be that that uh this is a a quite good approximation and questions i