diff --git a/CHANGES.md b/CHANGES.md
index da7564f3..ab1f6263 100644
--- a/CHANGES.md
+++ b/CHANGES.md
@@ -1,4 +1,8 @@
 
+# 6.2.1 (2023-09-28)
+
+* allow GeoJSON `Feature` in `ImageData.get_coverage_array` method
+
 # 6.2.0 (2023-09-27)
 
 * allow area-weighted statistics by adding `coverage` option in `rio_tiler.utils.get_array_statistics`
diff --git a/rio_tiler/models.py b/rio_tiler/models.py
index f151b0fa..d74068c1 100644
--- a/rio_tiler/models.py
+++ b/rio_tiler/models.py
@@ -798,19 +798,26 @@ def get_coverage_array(
         shape_crs: CRS = WGS84_CRS,
         cover_scale: int = 10,
     ) -> NDArray[numpy.floating]:
+        """Post-process image data.
 
-        """
-        cover_scale: int, optional
-            Scale used when generating coverage estimates of each
-            raster cell by vector feature. Coverage is generated by
-            rasterizing the feature at a finer resolution than the raster then using a summation to aggregate
-            to the raster resolution and dividing by the square of cover_scale
-            to get coverage value for each cell. Increasing cover_scale
-            will increase the accuracy of coverage values; three orders
-            magnitude finer resolution (cover_scale=1000) is usually enough to
-            get coverage estimates with <1% error in individual edge cells coverage
-            estimates, though much smaller values (e.g., cover_scale=10) are often
-            sufficient (<10% error) and require less memory.
+        Args:
+            in_range (tuple): input min/max bounds value to rescale from.
+            out_dtype (str, optional): output datatype after rescaling. Defaults to `uint8`.
+            color_formula (str, optional): color-ops formula (see: https://github.com/vincentsarago/color-ops).
+            cover_scale (int, optional):
+                Scale used when generating coverage estimates of each
+                raster cell by vector feature. Coverage is generated by
+                rasterizing the feature at a finer resolution than the raster then using a summation to aggregate
+                to the raster resolution and dividing by the square of cover_scale
+                to get coverage value for each cell. Increasing cover_scale
+                will increase the accuracy of coverage values; three orders
+                magnitude finer resolution (cover_scale=1000) is usually enough to
+                get coverage estimates with <1% error in individual edge cells coverage
+                estimates, though much smaller values (e.g., cover_scale=10) are often
+                sufficient (<10% error) and require less memory.
+
+        Returns:
+            numpy.array: percent coverage.
 
         Note: code adapted from https://github.com/perrygeo/python-rasterstats/pull/136 by @sgoodm
 
@@ -818,6 +825,7 @@ def get_coverage_array(
         if self.crs != shape_crs:
             shape = transform_geom(shape_crs, self.crs, shape)
 
+        shape = shape.get("geometry", shape)
         cover_array = rasterize(
             [(shape, 1)],
             out_shape=(self.height * cover_scale, self.width * cover_scale),
diff --git a/tests/test_models.py b/tests/test_models.py
index aa02d761..2e200fa3 100644
--- a/tests/test_models.py
+++ b/tests/test_models.py
@@ -456,3 +456,6 @@ def test_imagedata_coverage():
     }
     coverage = im.get_coverage_array(poly)
     assert numpy.unique(coverage).tolist() == [0.25]
+
+    coverage = im.get_coverage_array({"type": "Feature", "geometry": poly})
+    assert numpy.unique(coverage).tolist() == [0.25]