Completely blind image quality assessment is the further development of no reference image quality assessment. And this kind quality assessment method is highly unsupervised, training free, and don’t rely on natural scene statistic model. This paper proposes a completely blind quality assessment method based on gray-scale fluctuations (GFQA). The new method uses a specific image primitive to analyze the image gray fluctuations. And the analysis result is used to assess the image quality and give the image quality assessment result directly. The experimental results show that the new method accords closely with human subjective judgments of diverse distorted images when tested on the large publicly available ‘LIVE’ Image Quality database.