Image Quality Assessment (IQA) is very important in many different applications. It is therefore not surprising that research into IQA has received extensive attention during the last three decades. Recent models in the field of IQA demonstrate strong performance on several standard IQA datasets. However, their reliance on computationally intensive deep learning architectures and/or complex calculations makes them unsuitable for resource-constrained systems such as embedded and mobile Systems. In this paper we propose a Full Reference (FR) IQA model, called DeepMSE, which is based on SqueezeNet for feature extraction and Mean square Error (MSE) for aggregation. Unlike existing FR-IQA models, the proposed model doesn't require training or tuning with Mean Opinion Scores (MOSs) , which helps mitigate the risk of overfitting. Additionally, our model reduces computational complexity while maintaining high performance, making it well-suited for deployment on mobile or edge devices. Experimental evaluations across large standard IQA datasets demonstrate the high performance of our model and its superiority over state-of-the-art methods in aligning with human visual perception, all while maintaining simplicity, compact size, and reduced complexity.