diff --git a/intrinsiccv/include/types.h b/intrinsiccv/include/types.h
index f6703c73891341c45fecd920d039a8d44c19ed21..a21bba91967ce7f98db3772be85b8bf36b0ff533 100644
--- a/intrinsiccv/include/types.h
+++ b/intrinsiccv/include/types.h
@@ -1,4 +1,4 @@
-// SPDX-FileCopyrightText: 2023 Arm Limited and/or its affiliates <open-source-office@arm.com>
+// SPDX-FileCopyrightText: 2023 - 2024 Arm Limited and/or its affiliates <open-source-office@arm.com>
 //
 // SPDX-License-Identifier: Apache-2.0
 
@@ -541,13 +541,14 @@ template <typename OperationType, typename... RowTypes>
 void zip_parallel_rows(OperationType &operation, Rectangle rect,
                        RowTypes... rows) INTRINSICCV_STREAMING_COMPATIBLE {
   for (size_t row_index = 0; row_index < rect.height(); row_index += 2) {
-    operation.process_row(rect.width(), rows.as_columns()...);
-    // Call pre-increment operator on all elements in the parameter pack.
-    ((++rows), ...);
     // Handle the last odd row in a special way.
     if (INTRINSICCV_UNLIKELY(row_index == (rect.height() - 1))) {
       ((rows.make_single_row(), ...));
     }
+
+    operation.process_row(rect.width(), rows.as_columns()...);
+    // Call pre-increment operator on all elements in the parameter pack.
+    ((++rows), ...);
   }
 }
 
diff --git a/test/api/test_yuv_to_rgb.cpp b/test/api/test_yuv_to_rgb.cpp
index 876e4dce59171780a48e65a24cdcf33d9ff06934..931695a9ab9e1be8c8230fcee9c9d9669f9a459e 100644
--- a/test/api/test_yuv_to_rgb.cpp
+++ b/test/api/test_yuv_to_rgb.cpp
@@ -16,34 +16,44 @@ class YuvTest final {
   template <typename F>
   void execute_scalar_test(F impl, bool is_nv21) {
     size_t scalar_path_width = 5;
-    execute_test(impl, scalar_path_width, is_nv21);
+    execute_test(impl, scalar_path_width, is_nv21, 0);
+    // Padding version
+    execute_test(impl, scalar_path_width, is_nv21,
+                 test::Options::vector_length());
   }
 
   template <typename F>
   void execute_vector_test(F impl, bool is_nv21) {
     size_t vector_path_width = (2 * test::Options::vector_lanes<uint8_t>()) - 3;
-    execute_test(impl, vector_path_width, is_nv21);
+    execute_test(impl, vector_path_width, is_nv21, 0);
+    // Padding version
+    execute_test(impl, vector_path_width, is_nv21,
+                 test::Options::vector_length());
   }
 
  private:
   template <typename F>
-  void execute_test(F impl, size_t logical_width, bool is_nv21) {
-    test::Array2D<uint8_t> input_y{logical_width, 7};
+  void execute_test(F impl, size_t logical_width, bool is_nv21,
+                    size_t padding) {
+    test::Array2D<uint8_t> input_y{logical_width, 5, padding};
     input_y.set(0, 0, {10, 20, 255, 199});
     input_y.set(1, 0, {1, 120, 0, 17});
     input_y.set(2, 0, {2, 3, 240, 228});
-    input_y.set(6, 0, {7, 11, 128, 129});
+    input_y.set(4, 0, {7, 11, 128, 129});
 
     // the width of the UV input must be even
-    test::Array2D<uint8_t> input_uv{__builtin_align_up(logical_width, 2), 3};
+    test::Array2D<uint8_t> input_uv{__builtin_align_up(logical_width, 2), 3,
+                                    padding};
     input_uv.set(0, 0, {100, 130, 255, 255});
     input_uv.set(1, 0, {0, 1, 3, 4});
     input_uv.set(2, 0, {7, 8, 9, 10});
 
-    test::Array2D<uint8_t> expected{logical_width * channel_number_, 6};
+    test::Array2D<uint8_t> expected{logical_width * channel_number_,
+                                    input_y.height(), padding};
     calculate_expected(input_y, input_uv, expected, is_nv21);
 
-    test::Array2D<uint8_t> actual{logical_width * channel_number_, 6};
+    test::Array2D<uint8_t> actual{logical_width * channel_number_,
+                                  input_y.height(), padding};
 
     impl(input_y.data(), input_y.stride(), input_uv.data(), input_uv.stride(),
          actual.data(), actual.stride(), expected.width() / channel_number_,