diff --git a/src/dsp/dsp.h b/src/dsp/dsp.h
index d197eeb6..98782c88 100644
--- a/src/dsp/dsp.h
+++ b/src/dsp/dsp.h
@@ -86,6 +86,11 @@ typedef int (*VP8QuantizeBlock)(int16_t in[16], int16_t out[16],
                                 int n, const struct VP8Matrix* const mtx);
 extern VP8QuantizeBlock VP8EncQuantizeBlock;
 
+// specific to 2nd transform:
+typedef int (*VP8QuantizeBlockWHT)(int16_t in[16], int16_t out[16],
+                                   const struct VP8Matrix* const mtx);
+extern VP8QuantizeBlockWHT VP8EncQuantizeBlockWHT;
+
 // Collect histogram for susceptibility calculation and accumulate in histo[].
 struct VP8Histogram;
 typedef void (*VP8CHisto)(const uint8_t* ref, const uint8_t* pred,
diff --git a/src/dsp/enc.c b/src/dsp/enc.c
index 3eeadbef..aa71068a 100644
--- a/src/dsp/enc.c
+++ b/src/dsp/enc.c
@@ -11,7 +11,9 @@
 //
 // Author: Skal (pascal.massimino@gmail.com)
 
+#include <assert.h>
 #include <stdlib.h>  // for abs()
+
 #include "./dsp.h"
 #include "../enc/vp8enci.h"
 
@@ -652,6 +654,31 @@ static int QuantizeBlock(int16_t in[16], int16_t out[16],
   return (last >= 0);
 }
 
+static int QuantizeBlockWHT(int16_t in[16], int16_t out[16],
+                            const VP8Matrix* const mtx) {
+  int n, last = -1;
+  for (n = 0; n < 16; ++n) {
+    const int j = kZigzag[n];
+    const int sign = (in[j] < 0);
+    const int coeff = sign ? -in[j] : in[j];
+    assert(mtx->sharpen_[j] == 0);
+    if (coeff > mtx->zthresh_[j]) {
+      const int Q = mtx->q_[j];
+      const int iQ = mtx->iq_[j];
+      const int B = mtx->bias_[j];
+      out[n] = QUANTDIV(coeff, iQ, B);
+      if (out[n] > MAX_LEVEL) out[n] = MAX_LEVEL;
+      if (sign) out[n] = -out[n];
+      in[j] = out[n] * Q;
+      if (out[n]) last = n;
+    } else {
+      out[n] = 0;
+      in[j] = 0;
+    }
+  }
+  return (last >= 0);
+}
+
 //------------------------------------------------------------------------------
 // Block copy
 
@@ -686,6 +713,7 @@ VP8Metric VP8SSE4x4;
 VP8WMetric VP8TDisto4x4;
 VP8WMetric VP8TDisto16x16;
 VP8QuantizeBlock VP8EncQuantizeBlock;
+VP8QuantizeBlockWHT VP8EncQuantizeBlockWHT;
 VP8BlockCopy VP8Copy4x4;
 
 extern void VP8EncDspInitSSE2(void);
@@ -710,6 +738,7 @@ void VP8EncDspInit(void) {
   VP8TDisto4x4 = Disto4x4;
   VP8TDisto16x16 = Disto16x16;
   VP8EncQuantizeBlock = QuantizeBlock;
+  VP8EncQuantizeBlockWHT = QuantizeBlockWHT;
   VP8Copy4x4 = Copy4x4;
 
   // If defined, use CPUInfo() to overwrite some pointers with faster versions.
diff --git a/src/dsp/enc_sse2.c b/src/dsp/enc_sse2.c
index acc71edb..e47b1089 100644
--- a/src/dsp/enc_sse2.c
+++ b/src/dsp/enc_sse2.c
@@ -930,6 +930,11 @@ static int QuantizeBlockSSE2(int16_t in[16], int16_t out[16],
   }
 }
 
+static int QuantizeBlockWHTSSE2(int16_t in[16], int16_t out[16],
+                                const VP8Matrix* const mtx) {
+  return QuantizeBlockSSE2(in, out, 0, mtx);
+}
+
 #endif   // WEBP_USE_SSE2
 
 //------------------------------------------------------------------------------
@@ -941,6 +946,7 @@ void VP8EncDspInitSSE2(void) {
 #if defined(WEBP_USE_SSE2)
   VP8CollectHistogram = CollectHistogramSSE2;
   VP8EncQuantizeBlock = QuantizeBlockSSE2;
+  VP8EncQuantizeBlockWHT = QuantizeBlockWHTSSE2;
   VP8ITransform = ITransformSSE2;
   VP8FTransform = FTransformSSE2;
   VP8FTransformWHT = FTransformWHTSSE2;
diff --git a/src/enc/quant.c b/src/enc/quant.c
index 3ad49499..75e097dd 100644
--- a/src/enc/quant.c
+++ b/src/enc/quant.c
@@ -723,7 +723,7 @@ static int ReconstructIntra16(VP8EncIterator* const it,
     VP8FTransform(src + VP8Scan[n], ref + VP8Scan[n], tmp[n]);
   }
   VP8FTransformWHT(tmp[0], dc_tmp);
-  nz |= VP8EncQuantizeBlock(dc_tmp, rd->y_dc_levels, 0, &dqm->y2_) << 24;
+  nz |= VP8EncQuantizeBlockWHT(dc_tmp, rd->y_dc_levels, &dqm->y2_) << 24;
 
   if (DO_TRELLIS_I16 && it->do_trellis_) {
     int x, y;