Bug fix again for color space conversion

Author: rageworx

res/resource.h

@@ -1,2 +1,2 @@
-#define APP_VERSION								0,1,4,16
-#define APP_VERSION_STR							"0.1.4.16"
+#define APP_VERSION								0,1,4,17
+#define APP_VERSION_STR							"0.1.4.17"

src/debugtool.cpp

@@ -96,12 +96,13 @@ bool convertF32toU8( ImgF32* src, ImgU8 &dst )
 
     if ( fMin < 0.f )
     {
-        printf( "Error @ convertF32toU8(), Min float under zero : %.2f\n", fMin );
+        printf( "Warning @ convertF32toU8(), Min float under zero : %.2f\n", 
+                fMin );
         fflush( stdout );
-        
-        fMin = 0.f;
     }
 
+    printf( "fMin:fMax=%.2f:%.2f", fMin, fMax );
+    
     #pragma omp parallel for
     for( unsigned cnt=0; cnt<srcsz; cnt++ )
     {
@@ -164,14 +165,17 @@ void saveImgYCbCr( void* img, const char* fnameprefix )
     char strFnMap[1024] = {0};
 
     // Write Y
+    printf("saveImgYCbCr(%s), Y:", fnameprefix ); fflush( stdout );
     sprintf( strFnMap, "%s_Y.png", fnameprefix );
     saveImgF32( &refimg->Y, strFnMap );
 
     // Write Cb
+    printf("Cb:" ); fflush( stdout );
     sprintf( strFnMap, "%s_Cb.png", fnameprefix );
     saveImgF32( &refimg->Cb, strFnMap );
 
     // Write Cr
+    printf("Cr:" ); fflush( stdout );
     sprintf( strFnMap, "%s_Cr.png", fnameprefix );
     saveImgF32( &refimg->Cr, strFnMap );
 

src/libsrcnn.cpp

@@ -192,9 +192,9 @@ void converImgU8toYCbCr( ImgU8 &src, ImgYCbCr &out )
     #pragma omp parallel for
     for( unsigned cnt=0; cnt<imgsz; cnt++ )
     {
-        float fR = src.buff[ ( cnt * src.depth ) + 0 ];
-        float fG = src.buff[ ( cnt * src.depth ) + 1 ];
-        float fB = src.buff[ ( cnt * src.depth ) + 2 ];
+        float fR = (float)src.buff[ ( cnt * src.depth ) + 0 ];
+        float fG = (float)src.buff[ ( cnt * src.depth ) + 1 ];
+        float fB = (float)src.buff[ ( cnt * src.depth ) + 2 ];
 
         // Y
         out.Y.buff[cnt] = ( 0.299f * fR ) + 
@@ -212,6 +212,7 @@ void converImgU8toYCbCr( ImgU8 &src, ImgYCbCr &out )
                            ( 0.5f * fR ) - 
                            ( 0.4187f * fG ) - 
                            ( 0.0813f * fB );
+                       
     }
 }
 
@@ -239,9 +240,9 @@ void convertImgF32x3toImgU8( ImgF32* src, ImgU8 &out )
         float fB  = MIN(255.f, fY + 113.f * fCb / 64.f );
 
         // Red -> Green -> Blue ...
-        out.buff[( cnt * 3 ) + 0] = (unsigned char)fR;
-        out.buff[( cnt * 3 ) + 1] = (unsigned char)fG;
-        out.buff[( cnt * 3 ) + 2] = (unsigned char)fB;
+        out.buff[( cnt * 3 ) + 0] = (unsigned char)MAX( 0.f, fR );
+        out.buff[( cnt * 3 ) + 1] = (unsigned char)MAX( 0.f, fG );
+        out.buff[( cnt * 3 ) + 2] = (unsigned char)MAX( 0.f, fB );
     }
 }
 
@@ -528,8 +529,11 @@ int DLL_PUBLIC ProcessSRCNN( const unsigned char* refbuff,
     discardImgYCbCr( imgYCbCr );
 
 #ifdef DEBUG
+    printf("rY:");
     saveImgF32( &imgResized[0], "resized_Y.png" );
+    printf("rCb:");
     saveImgF32( &imgResized[1], "resized_Cb.png" );
+    printf("rCr:");
     saveImgF32( &imgResized[2], "resized_Cr.png" );
 #endif
 
@@ -608,4 +612,4 @@ int DLL_PUBLIC ProcessSRCNN( const unsigned char* refbuff,
     }
 
     return -100;
-}
+}