In this paper, I present a flattening method with hard constraints that exploits Discrete Cosine Transforms (DCTs) to increase computational efficiency. It is a modification of the constrained Gauss-Newton flattening method Lomask and Guitton (2006) using an improved preconditioner. The preconditioner, an unconstrained flattening method that uses DCTs, was presented by Lomask and Fomel (2006). Instead of approximating the inverse of the Laplacian with the helical transform, the DCT is exploited to more accurately invert the Laplacian. The resulting algorithm converges faster than previous constrained flattening methods while the memory usage is similar. Here, I first review the constrained Gauss-Newton flattening method. I then demonstrate its use on a faulted 3D field data set from the Gulf of Mexico.