Search results

Hello all, I am trying to understand Markov Random Fields and endearing graphs for optimisation with graph cuts. I went to a lecture the other day at a local university and the speaker said something like (from my notes): "In a an MRF the so-called "cliques" are not ordered. One does not...

Hello everyone, I am using the bicubic kernel described here ( Bicubic interpolation - Wikipedia, the free encyclopedia ) to interpolate my image after applying some transformations. I an using the matrix kernel described here with a = -0.5. Now, what I also need to do is estimate the...

Hi, Thanks for the reply. In this case, I know that this would be a convex function. So, I just have to walk down the slope and the minima that I find would be the right one. So, in this sense the only task is to go down the hill somehow in a fast way. I was using a simple gradient descent...

Yes, you are right. Can you comment on the validity of using a LAX type numerical scheme (I am not clear on what this entails here) to solve this problem? I am guessing the LAX scheme is also not valid here as there is no time component but I should compute the derivatives and the hessian...

Hello, Thanks for the reply. So, the equation in my original post is set up as a gradient descent system and represents one iteration of the gradient descent. (\psi_{n+1}) is the current estimate. (\psi_{n}) is the previous estimate. (\alpha) is the step-size of the gradient descent. I...

Hi Daniel, Thanks for the reply. Yes, the system is iterative but has no time dependence. The equation is there in my original post. It is a convex reformulation of an old problem and the equation is a gradient descent type of formulation. The last term is the hessian. The equation is in my...

The problem is actually an image processing problem where this is a gradient descent solution to a convex function. For the CN bit, can't the step size of the GD be viewed as the time parameter as at each iteration one can see it as a flow towards the solution. Thanks, Luc

Hi everyone, I have a gradient descent problem of the following form: (\psi_{n+1}=\psi_{n}+\alpha(\nabla\psi_{n}*D^{2}\psi)) I am trying this on a 256x256 image grid where everything is spaced uniformly and dx and dy =1. I am using s step size of 0.5 using a normal gradient descent...