I've read some papers about that two subjects while I can not tell the difference between them. Any body can help me?

In POD the fluid variables are written as a linear combination of some global modes..these global modes are computed by taking snapshots of the time history data of the fluid variables, putting them in a matrix, forming a spatial correlation matrix and then computing the eigenvalues/eigenvectors of the corelation matrix. Alternatively you can computed the POD modes by performing an SVD on the matrix of snapshot data. SVD itself is just a way to decompose a matrix.