diff --git a/python/paddle_fl/mpc/examples/mean_normalize_demo/README.md b/python/paddle_fl/mpc/examples/mean_normalize_demo/README.md
index 6200d070cf920e91f41de76ab5f2c8600dbe69d6..5d590c6e2e74585201e773f9ff32d6a0c5e866be 100644
--- a/python/paddle_fl/mpc/examples/mean_normalize_demo/README.md
+++ b/python/paddle_fl/mpc/examples/mean_normalize_demo/README.md
@@ -11,8 +11,9 @@ Create a empty dir for data, and modify `data_path` in `process_data.py`,
 default dir path is `./data`.
 
 Then run the script with command `python prepare.py` to generate random data
-for demo. Otherwise generate your own data, move them to `data_path` and modify
-corresponding meta info in `prepare.py`.
+for demo, which is dumped by numpy and named `feature_data.{i}.npy` located
+in `data_path`. Otherwise generate your own data, move them to `data_path`,
+name as the same way, and  modify corresponding meta info in `prepare.py`.
 
 Encrypted data files of feature statstics would be generated and saved in
 `data_path` directory. Different suffix names are used for these files to
@@ -55,5 +56,9 @@ import process_data
 res = process_data.decrypt_data(prepare.data_path + 'result', (2, prepare.feat_width, ))
 ```
 
-Also, `verify.py` could be used to calculate error between direct plaintext
-numpy calculation and mpc mean normalize.
+Or use `decrypt_and_rescale.py` to decrypt, rescale the feature data which has
+been saved in `feature_data.{i}.npy`, and dump the normalized data to
+`normalized_data.{i}.npy` which is located in `data_path`.
+
+Also, `verify.py` could be used to calculate error of `f_range` and `f_mean`
+between direct plaintext numpy calculation and mpc mean normalize.
diff --git a/python/paddle_fl/mpc/examples/mean_normalize_demo/decrypt_and_rescale.py b/python/paddle_fl/mpc/examples/mean_normalize_demo/decrypt_and_rescale.py
new file mode 100644
index 0000000000000000000000000000000000000000..e257a0d7c7c41bb1e611ab3bb5e5a73ae00ebbc5
--- /dev/null
+++ b/python/paddle_fl/mpc/examples/mean_normalize_demo/decrypt_and_rescale.py
@@ -0,0 +1,35 @@
+# Copyright (c) 2020 PaddlePaddle Authors. All Rights Reserved.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+"""
+Decrypt and rescale for mean normalize demo.
+"""
+import sys
+import numpy as np
+import process_data
+import prepare
+
+data_path = prepare.data_path
+# 0 for f_range, 1 for f_mean
+# use decrypted global f_range and f_mean to rescaling local feature data
+res = process_data.decrypt_data(data_path + 'result', (2, prepare.feat_width, ))
+
+party = sys.argv[1]
+
+input = np.load(data_path + 'feature_data.' + party + '.npy')
+
+output = (input - res[1]) / res[0]
+
+np.save(data_path + 'normalized_data.' + party, output)
+
+