matplotlib
diff --git a/‎lib/matplotlib/image.py
Lines changed: 7 additions & 6 deletions b/‎lib/matplotlib/image.py
Lines changed: 7 additions & 6 deletions
diff --git a/‎lib/matplotlib/tests/baseline_images/test_image/imshow_masked_interpolation.pdf
-9 Bytes b/‎lib/matplotlib/tests/baseline_images/test_image/imshow_masked_interpolation.pdf
-9 Bytes
diff --git a/‎lib/matplotlib/tests/baseline_images/test_image/imshow_masked_interpolation.png
5 Bytes b/‎lib/matplotlib/tests/baseline_images/test_image/imshow_masked_interpolation.png
5 Bytes
@@ -367,10 +367,11 @@ def _make_image(self, A, in_bbox, out_bbox, clip_bbox, magnification=1.0,
                 # need to
 
                 # TODO slice input array first
-
-                # make a working array up here, re-use twice to save memory
-                working_array = np.empty(A.shape, dtype=np.float32)
-
+                inp_dtype = A.dtype
+                if inp_dtype.kind == 'f':
+                    scaled_dtype = A.dtype
+                else:
+                    scaled_dtype = np.float32
                 a_min = np.nanmin(A)
                 a_max = np.nanmax(A)
                 # scale the input data to [.1, .9].  The Agg
@@ -380,7 +381,7 @@ def _make_image(self, A, in_bbox, out_bbox, clip_bbox, magnification=1.0,
                 # over / under.
                 # This may introduce numeric instabilities in very broadly
                 # scaled data
-                A_scaled = working_array
+                A_scaled = np.empty(A.shape, dtype=scaled_dtype)
                 A_scaled[:] = A
                 A_scaled -= a_min
                 A_scaled /= ((a_max - a_min) / 0.8)
@@ -410,7 +411,7 @@ def _make_image(self, A, in_bbox, out_bbox, clip_bbox, magnification=1.0,
                 if isinstance(self.norm, mcolors.NoNorm):
                     A_resampled = A_resampled.astype(A.dtype)
 
-                mask = working_array
+                mask = np.empty(A.shape, dtype=np.float32)
                 if A.mask.shape == A.shape:
                     # this is the case of a nontrivial mask
                     mask[:] = np.where(A.mask, np.float32(np.nan),