Protect exposure.histogram from integer overflow

jni · jni · commit 879f8bb8a88e · 2014-11-24T23:59:43.000+11:00
Fixes #1228 . I also removed a min() calculation and prevented an image copy when offset is 0, which is most of the time.
diff --git a/skimage/exposure/exposure.py b/skimage/exposure/exposure.py
@@ -62,9 +62,18 @@ def histogram(image, nbins=256):
     # For integer types, histogramming with bincount is more efficient.
     if np.issubdtype(image.dtype, np.integer):
         offset = 0
-        if np.min(image) < 0:
-            offset = np.min(image)
-        hist = np.bincount(image.ravel() - offset)
+        image_min = np.min(image)
+        if image_min < 0:
+            offset = image_min
+            image_range = np.max(image).astype(np.int64) - image_min
+            # get smallest dtype that can hold both minimum and offset maximum
+            offset_dtype = np.promote_types(np.min_scalar_type(image_range),
+                                            np.min_scalar_type(image_min))
+            if image.dtype != offset_dtype:
+                # prevent overflow errors when offsetting
+                image = image.astype(offset_dtype)
+            image = image - offset
+        hist = np.bincount(image.ravel())
         bin_centers = np.arange(len(hist)) + offset
 
         # clip histogram to start with a non-zero bin
diff --git a/skimage/exposure/tests/test_exposure.py b/skimage/exposure/tests/test_exposure.py
@@ -12,6 +12,27 @@
 from skimage.util.dtype import dtype_range
 
 
+# Test integer histograms
+# =======================
+
+def test_negative_overflow():
+    im = np.array([-1, 127], dtype=np.int8)
+    frequencies, bin_centers = exposure.histogram(im)
+    assert_array_equal(bin_centers, np.arange(-1, 128))
+    assert frequencies[0] == 1
+    assert frequencies[-1] == 1
+    assert_array_equal(frequencies[1:-1], 0)
+
+
+def test_all_negative_image():
+    im = np.array([-128, -1], dtype=np.int8)
+    frequencies, bin_centers = exposure.histogram(im)
+    assert_array_equal(bin_centers, np.arange(-128, 0))
+    assert frequencies[0] == 1
+    assert frequencies[-1] == 1
+    assert_array_equal(frequencies[1:-1], 0)
+
+
 # Test histogram equalization
 # ===========================
 
