you best in my on the body got information with the one on you but still speak also uh the company so i would like to a a kind of my to work so so is the T G minor but a when you can use yeah nice problem of minutes before starting to what yeah we can start five it's early but don't want that for start that for once one the ways people coming in maybe few so what a problem but minutes for but one one question was really useful for a single all you here are present the string oh the any or consider here from the or uh_huh but here's for marketing what fraction of room rest of five db well the question is that i can use the reading in terms of like S it's all of using the more robust tree but to be the mouse trying to paul what a things that doing is thing to think these of like guess there are only a you can be used feedback to get results way or i this base or remote start so right these i present proxy the method like in that i have divided my presentation for by first of it you you of hmmm you of their scope i i with that and you know you convex function minimize on the complex modeling right and of so that a that my me then um of different the uh the they are either with a one of the most competitive math based on a great for section and hmmm one on my proximity a technique based on proximity operator computing i project find a yeah present some results as a as the issue of of but it just i would that estimated very at like in uh the am so let's start with that this of the disparity given to uh to images the in from two different point of you we can see that the the so that that and that's is is presented by court why for example and uh uh is and the right image with uh X with different so we will uh not as by right my we define a is simply and this is between these to me so you will be uh this is a T and X prime why might white or or or of its a rectified so why Y but it is a is to be a oh problem to estimate a is is the main study so we ate to fine the corresponding pixel and the last i the state of the art the can find might yeah approach but in that we can uh can not mentioned feature matching approach where a with ace yeah curves and segments from the images also global methods as the dynamic programming and evaluation an approach where they are mean you at they minimize a global the energy from also some fines that use a of the cross or local correlation we those as a normalized cross-correlation correlation in want to follow we will a light by less a variational methods where our cost function would be uh and then are measure between the left image and that right image compensated by the despite its you and the we will use and then are or five that is the proper a lower semi-continuous convex function and the sum of the as this or will be on them on our limits it's the court but is this function is convex with respect to the disparity you so to avoid a non-convex optimize addition a problem we will uh suppose that the uh right image compensated by a by to you is footage we will consider the first taylor expansion of this not than the two and uh uh a uh a and fact we a compute this first or expansion and the first to make you but so i right image compensated by a but to you would be it well to the right image compensated by a and the of value you by mine is you might is to bottom to light by the or something to do and and of the disparity compensated right so to simplify the notation we will consider as D is the gradient R is the right image compensated by a it's paid to you bought class you bar T as the left so our cost function will be at most of five which is convex function function of the you minus R the some always is a on the support them at but as a with few it uh if we aim to search that's very to you that minimize i would criterion and this case that this problem is and it's pose problem because we can find uh a and and a finite in fine the sole use some of this that convex so we will allow uh and some convex constraint modeling prior knowledge and observe that that a of of our is to me that mean we will present our problem as a set estimating set theoretic estimating a problem that's mean we with search i would this to you that minimize our criterion J and to don't to then there's section of different constraint that we also there so what's can be these constraint it can be done and looks about the uh minimum and maximum amplitude of the disparity you or all that are also it's can be the uh uh uh you can produce an upper bound on the total variation of i would spend at mouth and this constraint to you get not does that that it's a piece there of the discontinuities and smooth some was it almost in is ideas or or can be a cost with that constraint we can introduce use an upper bound also on the norm and one norm of the weight of coefficient wavelet coefficients of our despite so to summarise we are minimizing a compact criterion J under different convex constraint as i said before that the sub the get projects is one method it's would used and uh before that two thousand and nine by me that and the uh use a fact two so this problem it's that convex what front to uh because of the algorithm that used and also they were lies to add convex "'cause" the convex there which is as far you might you box way so they they an is that what that the can a convex uh criterion and also they yeah uh so we use subgradient projections of the this to you on the different convex constraint that they can there this this minimizing is a on the uh number of close and are yeah uh and the two images left and right one that's mean we can that here just the peaks still that up here in the last and that i not though we will so our problem that is a convex function J and the different combat a that we consider but we are not hold lies to use just sickly convex quadratic function we have a a great flexibility and that's source of this criterion so to introduce i would want as the definition of that the proximity operator first if you have a points why that's you wanna project it on a convex set C uh we can find this projection by minimizing the come at the problem had to get a function of this constraint plus the uh quite that take distance between a and what so if we saw but this problem we can find the projection of the point why on the convex constraint for a place them to get a function by and that and that but three function F so our problem would be minimizing the function F plus the the distance between a and what this all use of this minimization problem will be the proximity operator of the function F at point point just to uh uh uh rewrite our problem and different way so we are minimizing our convex function J on the different convex comp constraint that can be modelled it it's by a line or operator and time the minimization is on the non on clothes and an are yeah and use the the error measure of five which is convex a function of to you minus R how we can solve that use a a to here i will present but are are good as a uh a which have especially specially will be a the P B X eight plus and yeah i go to them which i low as to minimize this context can at a complex function uh was some closed convex constraint C i uh just if we are able to compute the proximity operator of the criterion J and the project directory would what estimate on different constraints C i so our problem will be solved so simple it X a plus algorithm so we can a our algorithm was some way uh related to the uh convex constraint that you have and also go a a positive constant related to the of does zero point to the cost function that we have we initiate our our mattresses this and we compute the factor you which is and then G of different up a or operator that so uh we have a D complex cost as we see in that in each iteration and our there is a we project each month is on to convex a and also we compute the proximity operator of the criterion J divided by calm so we have our just that are i and i would this but T U by relaxation parameter lump so the most important than the side them that's we compute the proximity operator of the criterion J and we have X it's form for different convex a a cost function and the guy died projection on that the D convex a constraint that the i is not so uh so so to uh to improve the performance of this and good as a mall present some results the first one a corresponding to the peer what it off he how we present the left image and the related to going to tools and i present some generated disparity map using different method to can but to compare it to our method the first one is the block despite the estimation this a one that that i presented before using got to the what that the convex function based on sub granting projection and the our but the X a plus and go to them and we choose that of the and one norm and a a and that are much in our would cost of from also we compute the psnr between that is generated this to my and the ground truth as we see that our method gives the best result and some important what improvement and the disparity map that you have and of the results on the teddy pair so the last image it's and the ground the true and the generated disparity map using the same method oh i will have a real be the B D easy to it's of block that despite the estimation and the L two norm that is strictly convex function and and one norm based on the X eight but i as an application of the uh estimated disparity map will present uh that's that uh its application and stating image coding as we know that and uh to compare the method with an but uh that you image coding a we um codes first uh the last and the right image independently and here we apply the same at loss for on the left and right image using find three we have that like possible many works are done in a stereo image coding using john coding scheme uh based on a compensation in uh but the disparity a compensation this right image so at consist of code one image you which choose the right one and a residual image here here uh can define with the difference between the left image and the right image compensated by the despite its you and all the the the disparity was scored so we have to image to to the right the is it image and that is but and you were compared our we you method the block that but the if estimation that is by estimation using strictly a complex what the tick function an and one or based on the X eight plus and to the resulting wavelet coefficients are input in uh i i are encoded use it uh using a you back to thousand and also as a dense fields are and that the by applying a what be the compositions for the by an entropy coding with at point to six four so uh oh here we present the psnr of D reconstructed images left and the right one versus the bit error rate as we see that the it depends scheme uh give is of the the last P high and a couple states compensation this by the estimation here are give the best to all that to me in that john coding scheme uh give that uh the great yes and and you don't the L T P S and are given by our method presented by the red curve does this by the estimation using the and one norm and and then that the estimation do think this strictly convex function with it's which is whether T and a lot that estimation as i said before that our criterion J is complex and we have am a great flexibility and the choice of our i don't mess so we can use this so but to uh uh the case one our image images present some noise change that's me oh if you have a a a a for example so that paper oh the we the guide that it uh the court it or P with all that people are so we can use and one norm or or when we have a what's and noise we can use that this could but this that's for example because of when we have a a noise uh we we can get the prove that uh could but this does that can be used to uh to make a less this effect of patient so here we compare the at sort or to uh to D X i plus using and one or when we have a a so that paper noise changes and you can see that the uh and i where is all the the uh noise as the we don't have a is that in fact and that and when we have a a a noise the uh disparity map but present morse smooth are yeah and the P of this company D used for the objects that in that present in the uh a and the that disparity image and the some uh very numerical result using the snr and absolute you'd error measure or are given and that's prove that our method gives a uh are provide an a create that's uh map to conclude uh i would like to uh some right so we form our of our problem as a convex function that we are minimizing under different convex can and uh we have a uh are at that is to choose the error measure as we want so that we have various criteria and also just if we are able to compute the proximity operator of the criterion and also a make direct projections of the estimate on different convex constraint so we can run the algorithm simply and also i present two applications and the presence of noise uh when and the left and the right image and also on the is a uh and state you image coding a what i'm working on a this time is that to a a a uh i is then extending my yeah my and good as them to the case when i have a nation variation and also i would like to applied to a for the colour image that is that thank you for that then still have a questions it's really here but so that projections to compute or or a non be or numerically and the better remote yeah so what was that the traditional load of the scheme as compared to the mid those that are present uh competing to the of getting projection you have a an the other and yeah uh a fact in the lot by the estimation uh so log and the fact that based on a block disparity is so they uh compute block and the there are but in the in uh to get in projection that are obliged to compute the this again at an operator or or the uh so when we use when that use for example of the of to variation there are a like to compute the sub gradient difference of of the total variation and applied the projection the complex but no no method in fact so we use the uh a applied the projection of the operator on the comp or what loss in computational complexity how much computation how much time does it big run these different all tomatoes all comparable the what for and fact we a plot we are caught a we program this algorithm on matlab of software and now we are applying and do you P G P U because uh as we see that and the algorithm we have a the ability to light a a set of a D projection we can uh projected that the but and but i don't way and also the uh proximity operator can be uh computed and but ah so we are waiting the fact the results of because uh the work at the always the a that are working on of the questions you uh question uh and fact that we are going to the uh sub getting projection technique and it was proved that two thousand and nine at the best that technique get pop to the other methods so we are made of our uh given there is not better than the best techniques so oh the question but at let's thing we once again i