A methodology is presented to reduce the amount of coupled power for co-located multiple-input-multiple-output (MIMO) radar waveforms via convex optimization. Specifically, a set of predefined waveforms is desired to be implemented on a given uniform linear array (ULA) structure, where the coupling behavior for the array is given via the array scattering matrix. The predefined waveforms are assumed to have coupled power greater than the desired-10 dB power return loss threshold on all or some of the antenna elements throughout the MIMO transmission. This assumption is typically true for MIMO emissions that place a significant amount of energy near endfire when the array is impedance matched at bore-sight. Here, we consider two instantiations of Doppler-division multiple access (DDMA) MIMO waveforms: a monotonically increasing Doppler frequency assignment across the ULA, and a random frequency assignment. The DDMA waveforms are altered to limit the energy coupling between the antenna elements while minimizing the change (or maximizing the similarity) to the original set of waveforms. The altered set of waveforms is generated via minimization of a Quadratically-Constrained Quadratic Programming (QCQP) convex optimization problem formulation. The resulting emissions follow intuition where the emission near boresight is largely unchanged from the original while the maximum change happens near endfire.