Super-resolution (SR) is a well-studied problem in signal processing, particularly with regard to image and video applications. SR techniques are useful because unlike simple interpolation, they create a high-resolution signal from a low-resolution input by generating new information that was not previously present. A growing body of research shows progress in development of SR techniques using dictionary learning. We propose a method for SR of musical signals through matrix factorization and use of a known musical dictionary. The approximate matching pursuit (AMP) algorithm is used to query the dictionary and perform factorization, making the overall process efficient and scalable. By approximating the frequency information of a low-resolution input spectrogram as the linear combination of entries in the musical dictionary, we are able to closely match the input signal's missing high frequency information and thereby create a high-resolution output signal.