Abstract: The release and sharing of medical image big data can effectively improve the quality of medical services. Medical images contain sensitive information of patients, in order to prevent the disclosure of patient privacy during the sharing of medical image data, it is necessary to protect the privacy of medical images. As a security mechanism with strict mathematical definition and proof of privacy protection strength, differential privacy has been widely used in image data privacy protection. To achieve the privacy protection for medical image big data, this paper proposes a differential privacy scheme for medical image big data with wavelet multi-resolution analysis. This scheme designs the medical image data format on the Hadoop platform and designs the differential privacy protection algorithm of medical image big data based on the MapReduce framework. Existing image differential privacy methods protect entire image data with a same privacy level without considering the different privacy requirements of different data. To solve this problem, combining with the wavelet transform technology commonly used in medical image processing, this paper proposes a privacy budget allocation algorithm based on wavelet multi-resolution analysis. The algorithm measures the importance and privacy requirement of different wavelet subbands in wavelet domain, and allocates privacy budget according to the privacy requirement of each wavelet subband. Finally, this paper proposes a pixel differential disturbance algorithm, which disturbs every pixel based on differential privacy exponential mechanism. The experimental results show that the proposed scheme can implement differential privacy protection according to the privacy protection requirements of each wavelet subband. Under the same privacy budget, the image visual effect of this scheme can be improved by up to 97.7% compared with the control scheme, and the image classification effect can be improved by up to 87.2%. The performance experiment on the big data platform shows the proposed scheme can realize efficient differential privacy protection of medical image big data.