Bug fixed to copying C-Layer3 to Y, version updated to 0.1.6.23

Author: rageworx

README.md

@@ -34,12 +34,14 @@ This is an open source project from original of this:
 * MacOSX ( clang, llvm )
 
 ## Latest Changes
+### Verison 0.1.6.23
+* Fixed a small bug of wrong internal copying size.
+
+## Previous Changes
 ### Verison 0.1.6.22
 * Fixed bug of original source (ShuWang's SRCNN).
    - Use last layer (3) to Y channel at last construction.
 * Changed ProcessSRCNN() method to get optional convolutional result.
-
-## Previous Changes
 ### Verison 0.1.6.20
 * Fixed memory leak after convolution55.
 * Changed ProcessSRCNN() method to get convolutional gray.

res/resource.h

@@ -1,2 +1,2 @@
-#define APP_VERSION								0,1,6,22
-#define APP_VERSION_STR							"0.1.6.22"
+#define APP_VERSION								0,1,6,23
+#define APP_VERSION_STR							"0.1.6.23"

src/libsrcnn.cpp

@@ -651,7 +651,7 @@ int DLL_PUBLIC ProcessSRCNN( const unsigned char* refbuff,
     // ---------------------------------------------------------
 	// Copy Third layer result to Y channel.
 
-	unsigned convsz = imgResized[0].width * imgResized[0].height;
+	unsigned convsz = imgResized[0].width * imgResized[0].height * sizeof( float );
 	memcpy( imgResized[0].buff, imgConv3.buff, convsz );
 
     /* Convert the image from YCrCb to RGB Space */