   TVPrimDual.m Runs the Newton iteration for the primal-dual 
                problem.
   Cmult.m      Multiplies a vector by the C matrix, the matrix
                in (3.5) for the primal-dual problem.
   Dmult.m      Multiplies a vector by D (del) or D' (grad). 
   formE.m      Computes a diagonal matrix with the norm of the
                gradient of u at each gridpoint, and the inverse
                of the matrix.
   formF.m      Computes a block diagonal matrix F, defined
                by (I - w grad-u' / | grad-u |) in (3.5).
   DualStep.m   Evaluates (3.6) for delta_w.
   fcn.m        Evaluates the function in (2.8) and its gradient in
                (2.9) for the line search cvsrch and other purposes.
   findsd.m     Solves (5.3) for largest feasible step.

   cvsrch.m     Computes a (primal variable) step length via line search.
   cstep.m      Used by cvsrch.
   paramtest.m  Finds the initial regularization parameter using the 
		discrepancy principle	