diff --git a/requests/utils.py b/requests/utils.py
index 118e7e1b..44b3e016 100644
--- a/requests/utils.py
+++ b/requests/utils.py
@@ -453,20 +453,12 @@ def _parse_content_type_header(header):
     :return: tuple containing content type and dictionary of
          parameters
     """
-    if not header:
-        return None
-    # append delimiter on end to ensure at least two elements when split by ';'
-    header += ';'
-    # split content type's main value from params
-    tokens = header.split(';', 1)
-    content_type_index = 0
-    params_index = 1
 
-    content_type = tokens[content_type_index].strip()
-    params = tokens[params_index]
-    params_dict = dict()
+    tokens = header.split(';')
+    content_type, params = tokens[0].strip(), tokens[1:]
+    params_dict = {}  # Using dict is actually slower than a dictionary literal. Weird but tru
 
-    for param in params.split(';'):
+    for param in params:
         if param and not param.isspace():
             param = param.strip()
             key, value = param, True
@@ -476,6 +468,7 @@ def _parse_content_type_header(header):
             params_dict[key] = value
     return content_type, params_dict
 
+
 def get_encoding_from_headers(headers):
     """Returns encodings from given HTTP Header Dict.
 
diff --git a/tests/test_utils.py b/tests/test_utils.py
index e734b8f8..f89d15aa 100644
--- a/tests/test_utils.py
+++ b/tests/test_utils.py
@@ -472,14 +472,6 @@ def test_parse_dict_header(value, expected):
 
 @pytest.mark.parametrize(
     'value, expected', (
-        (
-            None,
-            None
-        ),
-(
-            '',
-            None
-        ),
         (
             'application/xml',
             ('application/xml', dict())