The paper aims to propose a real distortion IQA method based on image underlying features and high-level semantics in view of the fact that the existing non-reference IQA methods can not accurately evaluate the quality of the real distortion image. Firstly, k-means clustering algorithm was performed according to the underlying feature index of the image Secondly, the deep convolutional neural network (DCNN) was used to extract the first-grade high-level semantics in each group. Then, second-grade high-level semantics that can provide better representation of image features were obtained by performing multiple statistical functions on first-grade high-level semantics. Besides, we established an effective high-capacity regressor with high-level semantics and subjective mean opinion scores (MOS) values of the human eyes. The experimental results showed that the proposed model on the KonIQ-10k image database can predict the quality score effectively and achieve a high consistency with the corresponding MOS value. Spearman order correlation coefficient (SROCC) and Kendall order correlation coefficient (KROCC) can reach 0.95 and 0.97, respectively. The proposed method can quickly and accurately evaluate the quality of real distorted image.