timhoffm
diff --git a/‎lib/matplotlib/image.py
Copy file name to clipboardExpand all lines: lib/matplotlib/image.py
+17-15Lines changed: 17 additions & 15 deletions b/‎lib/matplotlib/image.py
Copy file name to clipboardExpand all lines: lib/matplotlib/image.py
+17-15Lines changed: 17 additions & 15 deletions
diff --git a/‎lib/matplotlib/tests/test_image.py
Copy file name to clipboardExpand all lines: lib/matplotlib/tests/test_image.py
+46Lines changed: 46 additions & 0 deletions b/‎lib/matplotlib/tests/test_image.py
Copy file name to clipboardExpand all lines: lib/matplotlib/tests/test_image.py
+46Lines changed: 46 additions & 0 deletions
@@ -501,25 +501,27 @@ def _make_image(self, A, in_bbox, out_bbox, clip_bbox, magnification=1.0,
                 if A.ndim == 2:  # interpolation_stage = 'rgba'
                     self.norm.autoscale_None(A)
                     A = self.to_rgba(A)
-
                 alpha = self.get_alpha()
-                if alpha is not None and np.ndim(alpha) > 0:
-                    output_alpha = _resample(self, alpha, out_shape, t, resample=True)
-                    output = _resample(  # resample rgb channels
-                        # alpha: float, should only be specified when alpha is a scalar
-                        self, _rgb_to_rgba(A[..., :3]), out_shape, t)
-                else:
-                    alpha = self._get_scalar_alpha()
-                    if A.shape[2] == 3:
-                        # No need to resample alpha or make a full array; NumPy will
-                        # expand this out and cast to uint8 if necessary when it's
-                        # assigned to the alpha channel below.
-                        output_alpha = (255 * alpha) if A.dtype == np.uint8 else alpha
+                if alpha is None:  # alpha parameter not specified
+                    if A.shape[2] == 3:  # image has no alpha channel
+                        output_alpha = 255 if A.dtype == np.uint8 else 1.0
                     else:
                         output_alpha = _resample(  # resample alpha channel
-                            self, A[..., 3], out_shape, t, alpha=alpha)
+                            self, A[..., 3], out_shape, t)
                     output = _resample(  # resample rgb channels
-                        self, _rgb_to_rgba(A[..., :3]), out_shape, t, alpha=alpha)
+                        self, _rgb_to_rgba(A[..., :3]), out_shape, t)
+                elif np.ndim(alpha) > 0:  # Array alpha
+                    # user-specified array alpha overrides the existing alpha channel
+                    output_alpha = _resample(self, alpha, out_shape, t)
+                    output = _resample(
+                        self, _rgb_to_rgba(A[..., :3]), out_shape, t)
+                else:  # Scalar alpha
+                    if A.shape[2] == 3:  # broadcast scalar alpha
+                        output_alpha = (255 * alpha) if A.dtype == np.uint8 else alpha
+                    else:  # or apply scalar alpha to existing alpha channel
+                        output_alpha = _resample(self, A[..., 3], out_shape, t) * alpha
+                    output = _resample(
+                        self, _rgb_to_rgba(A[..., :3]), out_shape, t)
                 output[..., 3] = output_alpha  # recombine rgb and alpha
 
             # output is now either a 2D array of normed (int or float) data
 
@@ -1711,3 +1711,49 @@ def test_resample_dtypes(dtype, ndim):
     axes_image = ax.imshow(data)
     # Before fix the following raises ValueError for some dtypes.
     axes_image.make_image(None)[0]
+
+
+@pytest.mark.parametrize('intp_stage', ('data', 'rgba'))
+@check_figures_equal()
+def test_interpolation_stage_rgba_respects_alpha_param(fig_test, fig_ref, intp_stage):
+    axs_tst = fig_test.subplots(2, 3)
+    axs_ref = fig_ref.subplots(2, 3)
+    ny, nx = 3, 3
+    scalar_alpha = 0.5
+    array_alpha = np.random.rand(ny, nx)
+
+    # When the image does not have an alpha channel, alpha should be specified
+    # by the user or default to 1.0
+    im_rgb = np.random.rand(ny, nx, 3)
+    im_concat_default_a = np.ones((ny, nx, 1))  # alpha defaults to 1.0
+    im_rgba = np.concatenate(  # combine rgb channels with array alpha
+        (im_rgb, array_alpha.reshape((ny, nx, 1))), axis=-1
+    )
+    axs_tst[0][0].imshow(im_rgb)
+    axs_ref[0][0].imshow(np.concatenate((im_rgb, im_concat_default_a), axis=-1))
+    axs_tst[0][1].imshow(im_rgb, interpolation_stage=intp_stage, alpha=scalar_alpha)
+    axs_ref[0][1].imshow(
+        np.concatenate(  # combine rgb channels with broadcasted scalar alpha
+            (im_rgb, scalar_alpha * im_concat_default_a), axis=-1
+        ), interpolation_stage=intp_stage
+    )
+    axs_tst[0][2].imshow(im_rgb, interpolation_stage=intp_stage, alpha=array_alpha)
+    axs_ref[0][2].imshow(im_rgba, interpolation_stage=intp_stage)
+
+    # When the image already has an alpha channel, multiply it by the
+    # scalar alpha param, or replace it by the array alpha param
+    axs_tst[1][0].imshow(im_rgba)
+    axs_ref[1][0].imshow(im_rgb, alpha=array_alpha)
+    axs_tst[1][1].imshow(im_rgba, interpolation_stage=intp_stage, alpha=scalar_alpha)
+    axs_ref[1][1].imshow(
+        np.concatenate(  # combine rgb channels with scaled array alpha
+            (im_rgb, scalar_alpha * array_alpha.reshape((ny, nx, 1))), axis=-1
+        ), interpolation_stage=intp_stage
+    )
+    new_array_alpha = np.random.rand(ny, nx)
+    axs_tst[1][2].imshow(im_rgba, interpolation_stage=intp_stage, alpha=new_array_alpha)
+    axs_ref[1][2].imshow(
+        np.concatenate(  # combine rgb channels with new array alpha
+            (im_rgb, new_array_alpha.reshape((ny, nx, 1))), axis=-1
+        ), interpolation_stage=intp_stage
+    )