As modern image quality assessment (IQA) algorithms gain widespread adoption, it is becoming increasingly important to achieve a balance between their computational efficiency and their quality prediction fidelity. One way to improve the efficiency of such algorithms to meet real-time constraints is to use simplistic models of visual perception, however such an approach has serious drawbacks in terms of poor quality predictions and limited robustness to changing distortions and viewing conditions. This thesis presents a better alternative, in which a best-in-class full-reference (FR) IQA algorithm is decomposed into a massively parallel algorithmic structure which allows for its efficient implementation on a set of graphics processing units (GPUs). The paper illustrates that an understanding of the GPU and CPU architectures, combined with detailed knowledge of the IQA algorithm, can lead to non-trivial speedups without compromising prediction accuracy. A single-GPU and a multi-GPU implementation showed a 24x and a 43x speedup, respectively, over the baseline CPU implementation.