X-space reconstructions suffer from blurring caused by the point spread function (PSF) of the Magnetic Particle Imaging (MPI) system. Here, we propose a deep learning method for deblurring x-space reconstructed images. Our proposed method learns an end-to-end mapping between the gridding-reconstructed collinear images from two partitions of a Lissajous trajectory and the underlying magnetic nanoparticle (MNP) distribution. This nonlinear mapping is learned using measurements from a coded calibration scene (CCS) to speed up the training process. Numerical experiments show that our learning-based method can successfully deblur x-space reconstructed images across a broad range of measurement signal-to-noise ratios (SNR) following training at a moderate SNR.