Image defocus clues are widely used for generating corresponding depth images because of their speed and convenience. However, when the defocus degree of the defocused image is exceptionally high or low, the depth information of the image is often lost, making the generated depth images unable to satisfy the actual user requirements. It is necessary to distinguish and separate these images. However, existing methods are not sufficiently accurate for distinguishing the blur degree of the defocused images in different scenes, and lack the unified standards. Hence, they cannot distinguish and separate the defocused image effectively. In this study, an automatic discrimination method for the defocused image based on the image gray ratio is proposed to solve this problem. First, by analyzing and using the gradient and frequency-domain features of different regions of the out-of-focus image, it can effectively distinguish the blurred and clear areas of the image. Second, the feature images are fused to obtain a fusion image. Because the defocus degrees of the blurred and clear regions of the out-of-focus image are somewhat different, the contrast between the gray values of the two parts in the fusion image is evident. The ratio is used as a criterion to assess the degree of defocusing. When the ratio of the defocused image exceeds the set threshold standard, the image does not satisfy the depth image generation conditions, and separation is performed automatically. Finally, by comparing the existing definition evaluation function with the results obtained using the proposed method for discriminating the degrees of blurring of different defocused blurred images in the same and different scenes, the proposed method can more accurately and rapidly distinguish the degrees of blurring of defocused images in the same and different scenes, effectively separate the unqualified defocused images, and improve the generation efficiency of depth images.