Image quality assessment (IQA) has become a hot issue in image processing, which aims to evaluate the quality of images automatically by a metric being consistent with subjective evaluations. Though deep learning based techniques have been applied to IQA models to achieve much progress, conventional IQA models that adopt deliberately designed features are still meaningful because by which we can learn what features are quality-aware and hence understand more about the way we improve image quality by enhancing quality-aware features. The first stage of conventional IQA model design is the quality-aware feature selection. Phase congruency (PC), as one of efficient saliency maps, which takes advantages of early visual feature, operates in frequency domain to measure local structure such as edges, corners, lines, etc. by computing the local amplitude and local energy in multiple scales. Conventional local PC feature is calculated with log-Gabor based filtration in several orientations, and is usually combined with other features in quality estimating. Recently, researchers suggested that spatially circular symmetric filters, such as gradient magnitude (GM) and Laplacian of Gaussian (LoG), are highly-efficient quality features that have been widely used in various IQA model designs. By considering the first-order derivative property of GM and the second-order derivative property of LoG, the two features are a suitable pairs for PC compositions while keeping the circular symmetric characteristic. Therefore, we use GM and LoG to construct a novel PC feature map and only use this feature map to build a full-reference (FR) IQA model, which is proved to be state-of-the-art on three benchmark databases. Furthermore, we replace the PC algorithm in the state-of-the-art FSIM metric with our PC computation, which achieves improved performance. This study suggests that the proposed circular symmetric PC feature is a high efficient quality feature and can be exclusively used in IQA model designs.