diff --git a/src/CMakeLists.txt b/src/CMakeLists.txt
index cb7570a02..b3ec5d754 100644
--- a/src/CMakeLists.txt
+++ b/src/CMakeLists.txt
@@ -31,6 +31,7 @@ set(ncnn_SRCS
     mat_pixel_drawing.cpp
     mat_pixel_resize.cpp
     mat_pixel_rotate.cpp
+    transform.cpp
     modelbin.cpp
     net.cpp
     option.cpp
@@ -743,6 +744,7 @@ if(NCNN_INSTALL_SDK)
         layer_shader_type.h
         layer_type.h
         mat.h
+        transform.h
         modelbin.h
         net.h
         option.h
diff --git a/src/transform.cpp b/src/transform.cpp
new file mode 100644
index 000000000..368c5d3d7
--- /dev/null
+++ b/src/transform.cpp
@@ -0,0 +1,159 @@
+// Tencent is pleased to support the open source community by making ncnn available.
+//
+// Copyright (C) 2020 THL A29 Limited, a Tencent company. All rights reserved.
+//
+// Licensed under the BSD 3-Clause License (the "License"); you may not use this file except
+// in compliance with the License. You may obtain a copy of the License at
+//
+// https://opensource.org/licenses/BSD-3-Clause
+//
+// 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.
+
+#include <math.h>
+
+#include "transform.h"
+
+namespace ncnn {
+
+namespace transform {
+
+Mat combine(const Mat& m1, const Mat& m2)
+{
+    Mat M(3, 2, sizeof(float));
+    float row3[] = {0, 0, 1};
+    float* p = M;
+    const float* p1 = m1;
+    const float* p2 = m2;
+    for (int row = 0; row < 2; ++row)
+    {
+        for (int col = 0; col < 3; ++col)
+        {
+            p[row * 3 + col] = p1[row * 3 + 0] * p2[0 * 3 + col] + p1[row * 3 + 1] * p2[1 * 3 + col] + p1[row * 3 + 2] * row3[col];
+        }
+    }
+    return M;
+}
+
+Mat rotation(const float degree, const float* center)
+{
+    int ndim = 2;
+    Mat M(3, 2, sizeof(float));
+    float* mat = M;
+    float rad = degree * 3.14159265358979323846 / 180.0;
+    float c = cos(rad);
+    float s = sin(rad);
+
+    M.fill(0);
+    mat[0] = c;
+    mat[1] = -s;
+    mat[3] = s;
+    mat[4] = c;
+
+    if (center)
+    {
+        for (int d = 0; d < ndim; d++)
+        {
+            mat[d * 3 + ndim] = center[d] - center[0] * mat[d * 3 + 0] - center[1] * mat[d * 3 + 1];
+        }
+    }
+
+    return M;
+}
+
+Mat shear(const float degree, float* center)
+{
+    int ndim = 2;
+    Mat M(3, 2, sizeof(float));
+    float* mat = M;
+    float rad = degree * 3.14159265358979323846 / 180.0;
+    float s = tan(rad);
+
+    M.fill(0);
+    mat[0] = 1;
+    mat[1] = s;
+    mat[3] = s;
+    mat[4] = 1;
+
+    if (center)
+    {
+        for (int d = 0; d < ndim; d++)
+        {
+            mat[d * 3 + ndim] = center[d] - center[0] * mat[d * 3 + 0] - center[1] * mat[d * 3 + 1];
+        }
+    }
+
+    return M;
+}
+
+Mat shear(const float* degrees, const float* center)
+{
+    int ndim = 2;
+    Mat M(3, 2, sizeof(float));
+    float* mat = M;
+    float rad1 = degrees[0] * 3.14159265358979323846 / 180.0;
+    float s1 = tan(rad1);
+    float rad2 = degrees[1] * 3.14159265358979323846 / 180.0;
+    float s2 = tan(rad2);
+
+    M.fill(0);
+    mat[0] = 1;
+    mat[1] = s1;
+    mat[3] = s2;
+    mat[4] = 1;
+
+    if (center)
+    {
+        for (int d = 0; d < ndim; d++)
+        {
+            mat[d * 3 + ndim] = center[d] - center[0] * mat[d * 3 + 0] - center[1] * mat[d * 3 + 1];
+        }
+    }
+
+    return M;
+}
+
+Mat scale(const float* scale, const float* center)
+{
+    int ndim = 2;
+    Mat M(3, 2, sizeof(float));
+    M.fill(0);
+    float* mat = M;
+
+    for (int d = 0; d < ndim; d++)
+    {
+        mat[d * 3 + d] = scale[d];
+    }
+    if (center)
+    {
+        for (int d = 0; d < ndim; d++)
+        {
+            mat[d * 3 + ndim] = center[d] * (1 - scale[d]);
+        }
+    }
+
+    return M;
+}
+
+Mat translation(const float* offsets)
+{
+    int ndim = 2;
+    Mat M(3, 2, sizeof(float));
+    float* mat = M;
+
+    M.fill(0);
+    mat[0] = 1;
+    mat[4] = 1;
+    for (int d = 0; d < ndim; d++)
+    {
+        mat[d * 3 + ndim] = offsets[d];
+    }
+
+    return M;
+}
+
+} // namespace transform
+
+} // namespace ncnn
diff --git a/src/transform.h b/src/transform.h
new file mode 100644
index 000000000..34ef819d5
--- /dev/null
+++ b/src/transform.h
@@ -0,0 +1,46 @@
+// Tencent is pleased to support the open source community by making ncnn available.
+//
+// Copyright (C) 2020 THL A29 Limited, a Tencent company. All rights reserved.
+//
+// Licensed under the BSD 3-Clause License (the "License"); you may not use this file except
+// in compliance with the License. You may obtain a copy of the License at
+//
+// https://opensource.org/licenses/BSD-3-Clause
+//
+// 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.
+
+#ifndef NCNN_TRANSFORM_H
+#define NCNN_TRANSFORM_H
+
+#include "mat.h"
+
+namespace ncnn {
+
+namespace transform {
+
+Mat combine(const Mat& m1, const Mat& m2);
+
+template<typename... Args>
+Mat combine(const Mat& m1, const Mat& m2, const Args&... mats)
+{
+    return combine(combine(m1, m2), mats...);
+}
+
+Mat rotation(const float degree, const float* center = NULL);
+
+Mat shear(const float degree, float* center = NULL);
+
+Mat shear(const float* degrees, const float* center = NULL);
+
+Mat scale(const float* scale, const float* center = NULL);
+
+Mat translation(const float* offsets);
+
+} // namespace transform
+
+} // namespace ncnn
+
+#endif // NCNN_TRANSFORM_H