diff --git a/rtengine/LUT.h b/rtengine/LUT.h
index ca4ef64ea..ad43ab471 100644
--- a/rtengine/LUT.h
+++ b/rtengine/LUT.h
@@ -145,7 +145,7 @@ public:
 	}
 	// use with float indices
 	T operator[](float index) {
-		int idx = floor(index);
+		int idx = (int)index;  // don't use floor! The difference in negative space is no problems here
 		if (((unsigned int)idx) > maxs) {
 			if (idx<0)
 			{
diff --git a/rtengine/dcrop.cc b/rtengine/dcrop.cc
index 717b3aff6..29b541eb5 100644
--- a/rtengine/dcrop.cc
+++ b/rtengine/dcrop.cc
@@ -182,9 +182,9 @@ void Crop::update (int todo) {
 	}
 
     // switch back to rgb
-    parent->ipf.lab2rgb (labnCrop, cropImg);
+    parent->ipf.lab2monitorRgb (labnCrop, cropImg);
 	
-	//parent->ipf.lab2rgb (laboCrop, cropImg);
+	//parent->ipf.lab2monitorRgb (laboCrop, cropImg);
 	
 	//cropImg = baseCrop->to8();
 	/*
diff --git a/rtengine/improccoordinator.cc b/rtengine/improccoordinator.cc
index fba4ce1d3..65e66b974 100644
--- a/rtengine/improccoordinator.cc
+++ b/rtengine/improccoordinator.cc
@@ -335,7 +335,7 @@ void ImProcCoordinator::updatePreviewImage (int todo, Crop* cropCall) {
         previmg->getMutex().lock();
         try
         {
-            ipf.lab2rgb (nprevl, previmg);
+            ipf.lab2monitorRgb (nprevl, previmg);
             delete workimg;
 			workimg = ipf.lab2rgb (nprevl, 0,0,pW,pH, params.icm.working);        
         }
diff --git a/rtengine/improcfun.cc b/rtengine/improcfun.cc
index 7f9c005e2..b30f7cd1a 100644
--- a/rtengine/improcfun.cc
+++ b/rtengine/improcfun.cc
@@ -157,8 +157,8 @@ void ImProcFunctions::firstAnalysis (Imagefloat* original, const ProcParams* par
 	if (monitor) {
         cmsHPROFILE iprof = iccStore->getXYZProfile ();       
         lcmsMutex->lock ();
-		monitorTransform = cmsCreateTransform (iprof, TYPE_RGB_FLT, monitor, TYPE_RGB_8, settings->colorimetricIntent,
-            cmsFLAGS_NOOPTIMIZE | cmsFLAGS_NOCACHE );  // NOCACHE is important for thread safety
+		monitorTransform = cmsCreateTransform (iprof, TYPE_RGB_16, monitor, TYPE_RGB_8, settings->colorimetricIntent,
+            cmsFLAGS_NOOPTIMIZE | cmsFLAGS_NOCACHE );  // NOCACHE is for thread safety, NOOPTIMIZE for precision
         lcmsMutex->unlock ();
 	}
 	
diff --git a/rtengine/improcfun.h b/rtengine/improcfun.h
index 7f1650c69..07bb0db06 100644
--- a/rtengine/improcfun.h
+++ b/rtengine/improcfun.h
@@ -121,7 +121,7 @@ class ImProcFunctions {
 		void colorCurve       (LabImage* lold, LabImage* lnew);
 		void sharpening       (LabImage* lab, float** buffer);
 		void transform        (Imagefloat* original, Imagefloat* transformed, int cx, int cy, int sx, int sy, int oW, int oH);
-		void lab2rgb          (LabImage* lab, Image8* image);
+		void lab2monitorRgb   (LabImage* lab, Image8* image);
 		void resize           (Image16* src, Image16* dst, float dScale);
 		void deconvsharpening (LabImage* lab, float** buffer);
 		void MLsharpen (LabImage* lab);// Manuel's clarity / sharpening
diff --git a/rtengine/iplab2rgb.cc b/rtengine/iplab2rgb.cc
index d30c34879..2820b9362 100644
--- a/rtengine/iplab2rgb.cc
+++ b/rtengine/iplab2rgb.cc
@@ -54,7 +54,7 @@ const double (*iwprof[])[3] = {sRGB_xyz, adobe_xyz, prophoto_xyz, widegamut_xyz,
 const char* wprofnames[] = {"sRGB", "Adobe RGB", "ProPhoto", "WideGamut", "BruceRGB", "Beta RGB", "BestRGB"};
 const int numprof = 7;
 
-void ImProcFunctions::lab2rgb (LabImage* lab, Image8* image) {
+void ImProcFunctions::lab2monitorRgb (LabImage* lab, Image8* image) {
 	//MyTime tBeg,tEnd;
  //   tBeg.set();
 	//gamutmap(lab);
@@ -64,7 +64,8 @@ void ImProcFunctions::lab2rgb (LabImage* lab, Image8* image) {
         // cmsDoTransform is relatively expensive
         #pragma omp parallel for
 		for (int i=0; i<lab->H; i++) {
-            float buffer[3*lab->W];
+            // pre-conversion to integer, since the output is 8 bit anyway, but LCMS is MUCH faster not converting from float
+            unsigned short buffer[3*lab->W];
 
             const int ix = i * 3 * lab->W;
             int iy = 0;
@@ -85,9 +86,9 @@ void ImProcFunctions::lab2rgb (LabImage* lab, Image8* image) {
 				y_ = f2xyz(fy);
 				z_ = f2xyz(fz)*D50z;
 
-                buffer[iy++] = CLIP01(x_);
-                buffer[iy++] = CLIP01(y_);
-                buffer[iy++] = CLIP01(z_);
+                buffer[iy++] = (unsigned short)CLIP(x_* CMAXVAL+0.5);
+                buffer[iy++] = (unsigned short)CLIP(y_* CMAXVAL+0.5);
+                buffer[iy++] = (unsigned short)CLIP(z_* CMAXVAL+0.5);
 			}
 
             cmsDoTransform (monitorTransform, buffer, image->data + ix, lab->W);
@@ -174,9 +175,9 @@ Image8* ImProcFunctions::lab2rgb (LabImage* lab, int cx, int cy, int cw, int ch,
 				float y_ = 65535.0 * f2xyz(fy);
 				float z_ = 65535.0 * f2xyz(fz)*D50z;
 
-                buffer[iy++] = CLIP((int)x_);
-                buffer[iy++] = CLIP((int)y_);
-                buffer[iy++] = CLIP((int)z_);
+                buffer[iy++] = CLIP((int)(x_+0.5));
+                buffer[iy++] = CLIP((int)(y_+0.5));
+                buffer[iy++] = CLIP((int)(z_+0.5));
             }
 
             cmsDoTransform (hTransform, buffer, image->data + ix, cw);
@@ -258,9 +259,9 @@ Image16* ImProcFunctions::lab2rgb16 (LabImage* lab, int cx, int cy, int cw, int
 				float y_ = 65535.0 * f2xyz(fy);
 				float z_ = 65535.0 * f2xyz(fz)*D50z;
 
-				xa[j-cx] = CLIP((int)x_);
-				ya[j-cx] = CLIP((int)y_);
-				za[j-cx] = CLIP((int)z_);
+				xa[j-cx] = CLIP((int)(x_+0.5));
+				ya[j-cx] = CLIP((int)(y_+0.5));
+				za[j-cx] = CLIP((int)(z_+0.5));
 			}
 		}
 
diff --git a/rtengine/rtthumbnail.cc b/rtengine/rtthumbnail.cc
index b2906c56b..5d033cff2 100644
--- a/rtengine/rtthumbnail.cc
+++ b/rtengine/rtthumbnail.cc
@@ -775,7 +775,7 @@ IImage8* Thumbnail::processImage (const procparams::ProcParams& params, int rhei
 
     // obtain final image
     Image8* readyImg = new Image8 (fw, fh);
-    ipf.lab2rgb (labView, readyImg);
+    ipf.lab2monitorRgb (labView, readyImg);
     delete labView;
     delete baseImg;