diff --git a/CHANGES.md b/CHANGES.md
index 73998ab..620ccc2 100644
--- a/CHANGES.md
+++ b/CHANGES.md
@@ -9,6 +9,7 @@ v1.7.0 - YYYY-MM-DD
 - Added support for basic compound stream filters for ASCII85Decode support
   (Issue #11)
 - Added support for LZWDecode filters (Issue #11)
+- Added support for Unicode filenames on Windows (Issue #18)
 - Added support for writing object streams (Issue #101)
 - Added support for GIF files (Issue #145)
 - Added support for WebP files (Issue #144)
diff --git a/pdfio-file.c b/pdfio-file.c
index 2d6e792..9e1aaf9 100644
--- a/pdfio-file.c
+++ b/pdfio-file.c
@@ -1515,6 +1515,87 @@ add_obj(pdfio_file_t   *pdf,		// I - PDF file
 }
 
 
+#if _WIN32
+//
+// '_pdfio_win32_open()' - Open or create a file.
+//
+// This function handles mapping of UTF-8 filenames to UTF-16 on Windows.
+//
+
+int					// O - File descriptor or -1 on error
+_pdfio_win32_open(const char *filename,	// I - UTF-8 filename
+		  int        oflag,	// I - Open flags
+		  int        mode)	// I - File permissions
+{
+  wchar_t	utf16name[MAXPATHLEN + 1],
+  					// UTF-16 filename
+		*utf16ptr;		// Pointer into UTF-16 filename
+  int		unich;			// Unicode character
+
+
+  // Convert the UTF-8 string to UTF-16...
+  utf16ptr = utf16name;
+  while (*filename && utf16ptr < (utf16name + sizeof(utf16name) / sizeof(utf16name[0]) - 2);)
+  {
+    if ((unich = *filename++) & 0x80)
+    {
+      if ((unich & 0xe0) == 0xc0 && (*filename & 0xc0) == 0x80)
+      {
+        // 2-byte UTF-8
+        unich = ((unich & 0x1f) << 6) | (*filename & 0x3f);
+        filename ++;
+      }
+      else if ((unich & 0xf0) == 0xe0 && (*filename & 0xc0) == 0x80 && (filename[1] & 0xc0) == 0x80)
+      {
+        // 3-byte UTF-8
+        unich = ((unich & 0x0f) << 12) | ((*filename & 0x3f) << 6) | (filename[1] & 0x3f);
+        filename += 2;
+      }
+      else if ((unich & 0xf8) == 0xf0 && (*filename & 0xc0) == 0x80 && (filename[1] & 0xc0) == 0x80 && (filename[2] & 0xc0) == 0x80)
+      {
+        // 4-byte UTF-8
+        unich = ((unich & 0x07) << 18) | ((*filename & 0x3f) << 12) | ((filename[1] & 0x3f) << 6) | (filename[2] & 0x3f);
+        filename += 3;
+      }
+      else
+      {
+        // Invalid UTF-8 char...
+        errno = EINVAL;
+        return (-1);
+      }
+    }
+
+    // Copy the unicode character...
+    if (unich > 0xffff)
+    {
+      // Two-word sequence...
+      *utf16ptr++ = 0xd800 | ((unich >> 10) & 0x03ff);
+      *utf16ptr++ = 0xdc00 | (unich 0x03ff);
+    }
+    else
+    {
+      // One-word...
+      *utf16ptr++ = unich;
+    }
+  }
+
+  *utf16ptr = '\0';
+
+  if (*filename)
+  {
+    // Filename too long...
+    errno = EINVAL;
+    return (-1);
+  }
+  else
+  {
+    // Pass on to _wopen...
+    return (_wopen(utf16name, oflag, mode));
+  }
+}
+#endif // WIN32
+
+
 //
 // 'compare_objmaps()' - Compare two object maps...
 //
diff --git a/pdfio-private.h b/pdfio-private.h
index 19b2007..94c7521 100644
--- a/pdfio-private.h
+++ b/pdfio-private.h
@@ -30,7 +30,7 @@
 #    define fileno	_fileno
 #    define lseek(f,o,w) (off_t)_lseek((f),(long)(o),(w))
 #    define mkdir(d,p)	_mkdir(d)
-#    define open	_open
+#    define open	_pdfio_win32_open
 #    define read(f,b,s)	_read((f),(b),(unsigned)(s))
 #    define rmdir	_rmdir
 #    define snprintf	_snprintf
@@ -412,6 +412,7 @@ extern size_t		_pdfio_strlcpy(char *dst, const char *src, size_t dstsize) _PDFIO
 extern double		_pdfio_strtod(pdfio_file_t *pdf, const char *s) _PDFIO_INTERNAL;
 extern void		_pdfio_utf16cpy(char *dst, const unsigned char *src, size_t srclen, size_t dstsize) _PDFIO_INTERNAL;
 extern ssize_t		_pdfio_vsnprintf(pdfio_file_t *pdf, char *buffer, size_t bufsize, const char *format, va_list ap) _PDFIO_INTERNAL;
+extern int		_pdfio_win32_open(const char *filename, int oflag, int mode) _PDFIO_INTERNAL;
 
 extern bool		_pdfioArrayDecrypt(pdfio_file_t *pdf, pdfio_obj_t *obj, pdfio_array_t *a, size_t depth) _PDFIO_INTERNAL;
 extern void		_pdfioArrayDebug(pdfio_array_t *a, FILE *fp) _PDFIO_INTERNAL;