diff --git a/aom_dsp/aom_dsp_rtcd_defs.pl b/aom_dsp/aom_dsp_rtcd_defs.pl
index f5c9b03..d672d77 100755
--- a/aom_dsp/aom_dsp_rtcd_defs.pl
+++ b/aom_dsp/aom_dsp_rtcd_defs.pl
@@ -1279,9 +1279,9 @@
   add_proto qw/unsigned int aom_mse8x8/, "const uint8_t *src_ptr, int  source_stride, const uint8_t *ref_ptr, int  recon_stride, unsigned int *sse";
 
   specialize qw/aom_mse16x16          sse2 avx2 neon msa/;
-  specialize qw/aom_mse16x8           sse2           msa/;
-  specialize qw/aom_mse8x16           sse2           msa/;
-  specialize qw/aom_mse8x8            sse2           msa/;
+  specialize qw/aom_mse16x8           sse2      neon msa/;
+  specialize qw/aom_mse8x16           sse2      neon msa/;
+  specialize qw/aom_mse8x8            sse2      neon msa/;
 
   if (aom_config("CONFIG_AV1_HIGHBITDEPTH") eq "yes") {
     foreach $bd (8, 10, 12) {
diff --git a/aom_dsp/arm/variance_neon.c b/aom_dsp/arm/variance_neon.c
index e1f4978..17d815c 100644
--- a/aom_dsp/arm/variance_neon.c
+++ b/aom_dsp/arm/variance_neon.c
@@ -401,9 +401,50 @@
   }
 }
 
-unsigned int aom_mse16x16_neon(const uint8_t *src, int src_stride,
-                               const uint8_t *ref, int ref_stride,
-                               unsigned int *sse) {
+static INLINE unsigned int mse8xh_neon(const uint8_t *src, int src_stride,
+                                       const uint8_t *ref, int ref_stride,
+                                       unsigned int *sse, int h) {
+  uint8x8_t s[2], r[2];
+  int16x4_t diff_lo[2], diff_hi[2];
+  uint16x8_t diff[2];
+  int32x4_t sse_s32[2] = { vdupq_n_s32(0), vdupq_n_s32(0) };
+
+  int i = 0;
+  do {
+    s[0] = vld1_u8(src);
+    src += src_stride;
+    s[1] = vld1_u8(src);
+    src += src_stride;
+    r[0] = vld1_u8(ref);
+    ref += ref_stride;
+    r[1] = vld1_u8(ref);
+    ref += ref_stride;
+
+    diff[0] = vsubl_u8(s[0], r[0]);
+    diff[1] = vsubl_u8(s[1], r[1]);
+
+    diff_lo[0] = vreinterpret_s16_u16(vget_low_u16(diff[0]));
+    diff_lo[1] = vreinterpret_s16_u16(vget_low_u16(diff[1]));
+    sse_s32[0] = vmlal_s16(sse_s32[0], diff_lo[0], diff_lo[0]);
+    sse_s32[1] = vmlal_s16(sse_s32[1], diff_lo[1], diff_lo[1]);
+
+    diff_hi[0] = vreinterpret_s16_u16(vget_high_u16(diff[0]));
+    diff_hi[1] = vreinterpret_s16_u16(vget_high_u16(diff[1]));
+    sse_s32[0] = vmlal_s16(sse_s32[0], diff_hi[0], diff_hi[0]);
+    sse_s32[1] = vmlal_s16(sse_s32[1], diff_hi[1], diff_hi[1]);
+
+    i += 2;
+  } while (i < h);
+
+  sse_s32[0] = vaddq_s32(sse_s32[0], sse_s32[1]);
+
+  *sse = horizontal_add_u32x4(vreinterpretq_u32_s32(sse_s32[0]));
+  return horizontal_add_u32x4(vreinterpretq_u32_s32(sse_s32[0]));
+}
+
+static INLINE unsigned int mse16xh_neon(const uint8_t *src, int src_stride,
+                                        const uint8_t *ref, int ref_stride,
+                                        unsigned int *sse, int h) {
   uint8x16_t s[2], r[2];
   int16x4_t diff_lo[4], diff_hi[4];
   uint16x8_t diff[4];
@@ -447,7 +488,7 @@
     sse_s32[3] = vmlal_s16(sse_s32[3], diff_hi[3], diff_hi[3]);
 
     i += 2;
-  } while (i < 16);
+  } while (i < h);
 
   sse_s32[0] = vaddq_s32(sse_s32[0], sse_s32[1]);
   sse_s32[2] = vaddq_s32(sse_s32[2], sse_s32[3]);
@@ -457,6 +498,21 @@
   return horizontal_add_u32x4(vreinterpretq_u32_s32(sse_s32[0]));
 }
 
+#define MSE_WXH_NEON(w, h)                                                 \
+  unsigned int aom_mse##w##x##h##_neon(const uint8_t *src, int src_stride, \
+                                       const uint8_t *ref, int ref_stride, \
+                                       unsigned int *sse) {                \
+    return mse##w##xh_neon(src, src_stride, ref, ref_stride, sse, h);      \
+  }
+
+MSE_WXH_NEON(8, 8)
+MSE_WXH_NEON(8, 16)
+
+MSE_WXH_NEON(16, 8)
+MSE_WXH_NEON(16, 16)
+
+#undef MSE_WXH_NEON
+
 unsigned int aom_get4x4sse_cs_neon(const uint8_t *src, int src_stride,
                                    const uint8_t *ref, int ref_stride) {
   uint8x8_t s[4], r[4];
diff --git a/test/variance_test.cc b/test/variance_test.cc
index d41b4be..46b87b1 100644
--- a/test/variance_test.cc
+++ b/test/variance_test.cc
@@ -2819,8 +2819,10 @@
                                                      &aom_get4x4sse_cs_neon)));
 
 INSTANTIATE_TEST_SUITE_P(NEON, AvxMseTest,
-                         ::testing::Values(MseParams(4, 4,
-                                                     &aom_mse16x16_neon)));
+                         ::testing::Values(MseParams(3, 3, &aom_mse8x8_neon),
+                                           MseParams(3, 4, &aom_mse8x16_neon),
+                                           MseParams(4, 4, &aom_mse16x16_neon),
+                                           MseParams(4, 3, &aom_mse16x8_neon)));
 
 const VarianceParams kArrayVariance_neon[] = {
   VarianceParams(7, 7, &aom_variance128x128_neon),