diff --git a/vp9/encoder/vp9_aq_cyclicrefresh.c b/vp9/encoder/vp9_aq_cyclicrefresh.c
index 98b273303..99e2c5a28 100644
--- a/vp9/encoder/vp9_aq_cyclicrefresh.c
+++ b/vp9/encoder/vp9_aq_cyclicrefresh.c
@@ -209,7 +209,7 @@ void vp9_cyclic_refresh_update_segment(VP9_COMP *const cpi,
   int is_skin = 0;
   if (refresh_this_block == 0 &&
       bsize <= BLOCK_16X16 &&
-      cpi->oxcf.content != VP9E_CONTENT_SCREEN) {
+      cpi->use_skin_detection) {
     is_skin = vp9_compute_skin_block(p[0].src.buf,
                                      p[1].src.buf,
                                      p[2].src.buf,
diff --git a/vp9/encoder/vp9_encodeframe.c b/vp9/encoder/vp9_encodeframe.c
index a67813e73..a2445b0c1 100644
--- a/vp9/encoder/vp9_encodeframe.c
+++ b/vp9/encoder/vp9_encodeframe.c
@@ -773,13 +773,14 @@ static int choose_partitioning(VP9_COMP *cpi,
     vp9_build_inter_predictors_sb(xd, mi_row, mi_col, BLOCK_64X64);
 
     // Check if most of the superblock is skin content, and if so, force split
-    // to 32x32. Avoid checking superblocks on/near boundary for high resoln
+    // to 32x32. Avoid checking superblocks on/near boundary and avoid low
+    // resolutons for now.
     // Note superblock may still pick 64X64 if y_sad is very small
     // (i.e., y_sad < cpi->vbp_threshold_sad) below. For now leave this as is.
     x->sb_is_skin = 0;
 #if !CONFIG_VP9_HIGHBITDEPTH
-    if (cpi->oxcf.content != VP9E_CONTENT_SCREEN && (low_res || (mi_col >= 8 &&
-        mi_col + 8 < cm->mi_cols && mi_row >= 8 && mi_row + 8 < cm->mi_rows))) {
+    if (cpi->use_skin_detection && !low_res && (mi_col >= 8 &&
+        mi_col + 8 < cm->mi_cols && mi_row >= 8 && mi_row + 8 < cm->mi_rows)) {
       int num_16x16_skin = 0;
       int num_16x16_nonskin = 0;
       uint8_t *ysignal = x->plane[0].src.buf;
diff --git a/vp9/encoder/vp9_encoder.c b/vp9/encoder/vp9_encoder.c
index 8c646938c..5341c9444 100644
--- a/vp9/encoder/vp9_encoder.c
+++ b/vp9/encoder/vp9_encoder.c
@@ -1671,6 +1671,7 @@ VP9_COMP *vp9_create_compressor(VP9EncoderConfig *oxcf,
   cpi->resize_state = 0;
   cpi->resize_avg_qp = 0;
   cpi->resize_buffer_underflow = 0;
+  cpi->use_skin_detection = 0;
   cpi->common.buffer_pool = pool;
 
   cpi->rc.high_source_sad = 0;
@@ -3130,7 +3131,7 @@ static void set_size_dependent_vars(VP9_COMP *cpi, int *q,
   if (oxcf->pass == 2 && cpi->sf.static_segmentation)
     configure_static_seg_features(cpi);
 
-#if CONFIG_VP9_POSTPROC
+#if CONFIG_VP9_POSTPROC && !(CONFIG_VP9_TEMPORAL_DENOISING)
   if (oxcf->noise_sensitivity > 0) {
     int l = 0;
     switch (oxcf->noise_sensitivity) {
@@ -3330,6 +3331,14 @@ static void encode_without_recode_loop(VP9_COMP *cpi,
   set_size_independent_vars(cpi);
   set_size_dependent_vars(cpi, &q, &bottom_index, &top_index);
 
+  if (cpi->oxcf.speed >= 5 &&
+      cpi->oxcf.pass == 0 &&
+      cpi->oxcf.rc_mode == VPX_CBR &&
+      cpi->oxcf.content != VP9E_CONTENT_SCREEN &&
+      cpi->oxcf.aq_mode == CYCLIC_REFRESH_AQ) {
+    cpi->use_skin_detection = 1;
+  }
+
   vp9_set_quantizer(cm, q);
   vp9_set_variance_partition_thresholds(cpi, q);
 
diff --git a/vp9/encoder/vp9_encoder.h b/vp9/encoder/vp9_encoder.h
index 76c36f4e5..838ceacd6 100644
--- a/vp9/encoder/vp9_encoder.h
+++ b/vp9/encoder/vp9_encoder.h
@@ -494,6 +494,8 @@ typedef struct VP9_COMP {
   int resize_buffer_underflow;
   int resize_count;
 
+  int use_skin_detection;
+
   NOISE_ESTIMATE noise_estimate;
 
   // VAR_BASED_PARTITION thresholds
diff --git a/vp9/encoder/vp9_noise_estimate.c b/vp9/encoder/vp9_noise_estimate.c
index b1a261986..e56cc9b01 100644
--- a/vp9/encoder/vp9_noise_estimate.c
+++ b/vp9/encoder/vp9_noise_estimate.c
@@ -162,7 +162,9 @@ void vp9_update_noise_estimate(VP9_COMP *const cpi) {
     for (mi_row = 0; mi_row < cm->mi_rows; mi_row++) {
       for (mi_col = 0; mi_col < cm->mi_cols; mi_col++) {
         // 16x16 blocks, 1/4 sample of frame.
-        if (mi_row % 4 == 0 && mi_col % 4 == 0) {
+        if (mi_row % 4 == 0 && mi_col % 4 == 0 &&
+            mi_row < cm->mi_rows - 1 &&
+            mi_col < cm->mi_cols - 1) {
           int bl_index = mi_row * cm->mi_cols + mi_col;
           int bl_index1 = bl_index + 1;
           int bl_index2 = bl_index + cm->mi_cols;