meaning yeah to go and the paper in the thing pretty with flat so i image the presentation divide it will be defined fall for as we we will give motivation then uh a very brief that but know then we you are going to review the the frame pretty each that's proposed for each the for the difficult to to for using this flat transform the proposed solution and modification the model selection after a going so it's some experimental results and then conclusion fusion the first question must be white to use that the transform when most of the in like to a specially if you do all standard five that's a one use plot and so uh the the fact that of using that and some since this a part of this used just a you in the block itself maybe to a that would be good read the shown the block yeah so besides that for the same reason we can exploit had to the sum of neighbour block that was before you know both of those the leads to a superior performance fine coding in objectivity in subject to and measure but that comes list the expense of for higher combat and this point the only as uh you might to coding standard that uses let but are song it i is J back so that was proposed but microsoft corporation it was formally known i those for and you from go it uh have proposed a a new interest frame prediction it not then and we propose here we propose a inter-frame prediction in the pixel domain much seem not what's then each thus no we'll go to a very brief okay uh you know that transforms at in block transform this you know is divided in block and that will be processed the in but different from block transform the uh this part will be bigger we have a a a a and this new row that will have in the time more samples then the original one that would call Z to a different from the traditional X log and is the but as the this uh as a that it will have been times more uh samples was this and will be called overlap set and we have a here for that the example for the overlapping factor of two we can see that half of the sample of but an important blocks in the last and the i will be used the of the traditional samples that were or a red to use it in the back B B and Q the matches is contained in the direct and inverse transform the sound will be held in the same way as in black dots on with the difference and uh and that the thing yeah will be different from the original one where that because the traditional criteria for perfect reconstruction is not no longer value if we want to have a perfect reconstruction would have to follow the this new criterion that is showed uh below little here okay that the point is that the blocks can not be a concern in that in for a construct this C no or a a does this know we are going to have to like late the contribution of several neighbour block and a that that i and not the difference from block to is that in the or the since we don't have all the name but neighbor block need that we have to care um a the tree no this see again uh the frame rate most so i i believe most of you is already familiar with this speech we well i wanted to remember that even though the it's not it's it's for what was proposed to be a video coding standard when used to encode in my you in my just it is a very efficient a image coding in one of the key features why i happens used the implementation of inter but each that consists on using the immediate neighbours but some neighbours of because of the encoder there close to that well to predict the a lot were and coding at this moment and in this way you just have to send the with here you can see that nine and prediction modes for a four by four in a by lot and we have additional for a prediction modes for a lots of sixteen by sixteen B oh Y what's the dish you could of using this is scheme in a lap but transforms and uh the for the of because to a good is C we it's than in it was at local not the words would have to have the a construction of the peaks as for use it in the prediction and we as we so we don't have the have construction before having all the neighbours we can not perform the the prediction of a a a a a a a a a broad we are we want to go be you the colours but uh uh the central out here the being the block we are going to in using a that are solved we would need the bigger and square but we have only this region here that a use is available for a constant we propose that instead of predicting the block itself that like to pretty the this is standard block in this way every single pixel so outside the bar far away from the as will be predicted for time for our overlapping factor of two yeah yeah the if we want to do this we would do is to be me sing half of the peaks as in the last one because and the black you hear have have not not a been uh but if should but we go to the way that jpeg that this it's not a six four in the block in which a about is a true putting a a a larger block we have a block of it by it is that the but in a model a block of sixty but sixteen and the blocks so oh oh O do in this way we can see that for the first block since the for a lot of the proof isn't mark or bob had already been called we need this case would have all the peaks is of variable for the prediction however when we go to the set and more we would go back to the the case we had before where the last i neighbours are not of a in the becomes worse because we have all rows half of the pixels in arts and uh and in the four we would only have i was more or not i don't know and see i D is is more corner to predict the the block to prove a little what we probe one of the things we propose is changing the order of encoding that this plots changing the had of those set and then the block in this way in the first we have the same conditions we had before so in the set and we go to the same question all the same case we had in the fourth level which means that uh we have just one corner in this case it got worse however i that are that we do didn't have the first the the left corner we again have so in this case we are going to ensure sure that half of the box will have all the pixels are available for the prediction and and the that have we have just this corner in this case we are going to propose to use just a D C a prediction that will consist of a pretty can all of the pieces as the average of the available peaks pre the process will want to you know a a much seem has done in in a you know the coder the the residual be couplet is difference i that that he's a big so and the prediction the after the that quantisation is a and inverse transform i can we have to remember that the it is you do you only if we did not have the quantisation is different from the original one and scenes we we have to mimic a perform adding the prediction what have to mimic eight this process in the prediction we simply we simply process it the prediction of before we need to the was you to to obtain that you have a we with the thing if we didn't have the prediction also also have tool change though the way that the mode so it's selected in we have maybe two mm make two wave of so like like no one is the reduction of the minimization of the prediction that's normally in measured by a sum of of bits a the difference which doesn't seem to be up but in this case because a is a big part as and the have construction of every C pixel so will be different depending on the position so we propose to weight the i and that this difference a going to need to the importance of the of of you of its simple uh and the proposed a weighting in is that we use in this work is given by these that would be the have constructed uh have that we would have if we had that much it just with one the rate distortion optimization can also be performed a but we have to remember that since we cannot not construct the plot the distortion has to be a and and measure in the transform domain and the a for the norm orthogonal transform we have to take a called the energy of a this is the uh one example of weighting matrix or the you generalized lab of the hot transform that at but scene well the few as well or thing by maximizing the coding gain for uh out of a grass is model where uh correlation factor of zero point nine five which is not to be a a good model for in and now we go to a and we will present uh with a so of for and the implementation that idea at at this point would was just to prove the concept that that the the of sounds can be used the again at we uh that but does ones could use a with friends good frame prediction so for simplicity reasons the than a dish was scaled denoted by we'll and because of this court the data coding but that was six at two eight by also we use a the general like to be a talk or not a song that would result for several reasons uh but the overlap you of overlapping factor of two maximise the number of lots that we can use the pretty the she she B T has a a good performance for this overlap factor and uh uh uh just every since the overlapping factor or or was of to we had a prediction block that once of sixteen by sixteen and different or we can implement any number of prediction modes at this more that's point only the four modes of a able in the H that six four were implement implemented when we are going to use let and so uh for comparison reasons also the traditional intra-frame prediction was implemented when using uh dct in this case all the nine modes were available and oh including them the mode we use we we is that the same approach to a propose for it's that the six for that means that means looking at the that an upper neighbour we but they the best uh the most probable mode and we can that in or the proposed of this scheme since the a or a neighbour didn't have that several models of prediction was seen play just to use to bit cold to to the half of the blocks in for the order have we don't have to encode anything "'cause" there is just one well and besides that that that application which means no additional like of flat but that's forms were also that you can see for uh are they imagine a better we can see that uh the proposed method that uses in blue are performed the transforms and then you a plus dct that we plan we can see also that most of taking in this case was that just by simple education of love but or some but but again our proposed method the uh use for the improvement in the coding this same result signal as a present by the first frame of the for which D C can give a bad in which again way our uh propose a method the or performed both in you inter a C T and and then can and we have for several other image in we she we can we can see you the result for the you the but not much or more comparing to in to plus D C T and and the right to we have uh lap buttons so we can see if here that we have a no okay uh now we go to the conclusions the results presented here sure that the entrance to addition could be adapted to be compatible with like also also we show that this propose a scheme are performs the application of lapped transform as well as the inter prediction with this at the you all tested in important to note that in our case we have just half of the block but the being predict and the new in have we have only for prediction modes we so if we implement that all the i'm not a different number of the results presented you could be for to improve we have a also it's a very preliminary results but this is going to imitation of the of the scheme presented the here in a a real which that's six for older we can see here that even though the gains present here i smaller we have a in all tested rate for the image by uh a few reasons why the gains are a smaller it it's that in this case where competing with not all not only the nine modes of fate by eight but then nine modes of four by four and the for of sixteen by sixteen and in our case we have only them for modes of uh that time limit in in eight by uh coding block also the one and a but may not be as well as a a a at that that two compared up to compress in a a lot but transforms transform utterance from coefficients as is in the dct i four sure where we are what what i've seen a i'll way of implementing infallible size prediction block as well what that thing out of a lot of factors to see we use what we lose in the prediction we can gain in the near future and and uh um we all i want to say you also that a station of this work has been a set than in i C two thousand and they extension to feed you once sept and will be presented nice so that can my presentation thank you and well but any questions from now i one quick question can you comment a computational complexity group okay just thing is the the suit uh in this case for times B the prediction will be done for time you can even though i did manage that that's see i can say that will be at least for time more a the in the part of prediction hand in first and the direct well but the point point the that or the power will be seen it to the it's that and question oh but was of speaker