Visual attention (VA) is an integral property of the human visual system. The deployment of a VA model can be beneficial for many image and video applications, such as, compression, transmission, and quality assessment. However, the design of a VA model is highly dependent on the subjective VA data used as ground truth and the application that the model is intended for. In this paper, we discuss two ways of obtaining subjective VA data that can subsequently be used to develop VA models; selective regions-of-interest and visual fixation patterns. The feasibility of both methods will be discussed, in particular, with respect to visual quality assessment.