High quality fundus photographs (FPs) are essential for clinicians to make accurate diagnosis of various ophthalmic diseases, including diabetic retinopathy, age-related macular degeneration, and glaucoma. Thus it becomes imperative that clinicians are presented with FPs, whose high diagnostic quality is assured. In this context, significant effort has been directed at developing automated tools that distinguish between high quality and low quality FPs. For this purpose, features suited to natural image quality assessment were traditionally employed even for diagnostic quality assessment of FPs. However, structure preserving features generated by deep scattering network (ScatNet) were recently reported to outperform aforementioned traditional features. In this paper, we demonstrate further improvement in performance by combining both the traditional features and ScatNet features. Importantly, additional improvement is witnessed when ScatNet features are computed in multicolor space. © 2021 IEEE.