Abstract : Over the past few years, there has been an increase in the creation and streaming of images by data providers. Among these images, those that feature faces have gained significant attention due to their applicability in entertainment and social media platforms. However, the large volume of images shared on these platforms poses significant challenges and demands extensive computing resources to facilitate effective data processing. Furthermore, images are susceptible to various distortions that can occur during processing, transmission, sharing, or when combined with other factors in real-life application scenarios. Hence, it is essential to evaluate image quality to ensure the acceptable delivery of content, even though some distorted images do not have access to their original version. In this paper, we introduce a framework that estimates the quality of images in real-time processing scenarios. Our quality assessment is achieved by combining a deep network with random forests. Furthermore, the evaluation of facial features is performed through a face alignment metric. The LIVE and TID2013 benchmark datasets, which were artificially distorted, were utilized for the experimental analysis. Our findings demonstrate that our proposed method surpasses the current state-of-the-art techniques, achieving a Pearson Correlation Coefficient (PCC) and Spearman Rank Order Correlation Coefficient (SROCC) of approximately 0.942 and 0.931, respectively, when compared to subjective human scores. Additionally, we were able to reduce processing time from 4.8ms to 1.8ms.