With the popularity of virtual reality, the distortion measurement plays an important role in the processing of panoramic image. The distortion of the panoramic image can make the statistical characteristics of the image change. However, most existing panoramic image quality assessment (IQA) methods ignore the statistical characteristics of panoramic image. Therefore, we propose to utilize adjacent pixels correlation (APC) feature to calculate the statistical characteristics and blindly assess panoramic image quality. Specifically, we find that the distortion can change the proportion of low-frequency information in the panoramic image. We calculate the probability distribution of adjacent pixel difference map using Markov chains to build APC feature, which captures the change in statistical characteristics. Ultimately, the APC feature are fed into support vector regression (SVR) for training and predicting the image quality score. Experiments show that our proposed algorithm is more accurate than the existing algorithms.