In this paper, we extract visual features from the spatial, angular, coupling, and projection domains considering the spatial-angular coupling of the plenoptic function to propose a nonreference method for assessing light-field image quality. Natural scene statistics (NSS) features of the central subaperture image are extracted in the spatial domain. In addition, macropixel and grayscale co-occurrence matrix (GLCM) features on epipolar plane images (EPI) are extracted in the angular and spatial-angular coupling domains, respectively, and local entropy statistical distribution characteristics of the refocusing maps are extracted in the projection domain. Then, multiple visual features are fused to form the visual feature vector of the light field, and support vector regression (SVR) is applied to train a scoring model. Thus, the light-field image quality assessment method based on multiple visual feature aggregation is established. Experimental results show that the proposed method shows consistency between the light-field score and subjective score.