Fixed an endianness bug in the PNG I/O functions which occurred when 16bit

grayscale PNGs were used.  libpng doesn't automatically convert from host
endianness to big endian as the PNG standard demands.  You have to explicitly
tell it to do this or it will write out the 16bit pixel data in host order
which result in the wrong thing happening on little endian chips.  Similarly
for reading PNGs.

dlib/image_loader/png_loader.cpp

@@ -12,6 +12,7 @@
 #include "png_loader.h"
 #include <png.h>
 #include "../string.h"
+#include "../byte_orderer.h"
 
 namespace dlib
 {
@@ -155,7 +156,11 @@ namespace dlib
         png_init_io( ld_->png_ptr_, fp );
         png_set_sig_bytes( ld_->png_ptr_, 8 );
         // flags force one byte per channel output
-        png_read_png( ld_->png_ptr_, ld_->info_ptr_, PNG_TRANSFORM_PACKING, NULL );
+        byte_orderer bo;
+        int png_transforms = PNG_TRANSFORM_PACKING;
+        if (bo.host_is_little_endian())
+            png_transforms |= PNG_TRANSFORM_SWAP_ENDIAN;
+        png_read_png( ld_->png_ptr_, ld_->info_ptr_, png_transforms, NULL );
         height_ = png_get_image_height( ld_->png_ptr_, ld_->info_ptr_ );
         width_ = png_get_image_width( ld_->png_ptr_, ld_->info_ptr_ );
         bit_depth_ = png_get_bit_depth( ld_->png_ptr_, ld_->info_ptr_ );

dlib/image_saver/save_png.cpp

@@ -9,6 +9,7 @@
 #include "save_png.h"
 #include <cstdio>
 #include <png.h>
+#include "../byte_orderer.h"
 
 namespace dlib
 {
@@ -95,7 +96,11 @@ namespace dlib
             png_init_io(png_ptr, fp);
 
 
-            const int png_transforms = PNG_TRANSFORM_IDENTITY;
+            int png_transforms = PNG_TRANSFORM_IDENTITY;
+            byte_orderer bo;
+            if (bo.host_is_little_endian())
+                png_transforms |= PNG_TRANSFORM_SWAP_ENDIAN;
+
             const long height = row_pointers.size();