The easiest way to do this is through some kind of 3rd party ActiveX such as LeadTools from www.leadtools.com on the web. This has the advantage of exposing image comments and other embedded information in a consistent fashion irrespective of the source image format.

From a low level point of view you don't really want to be decoding the stream markers to get at the COMment (0xFE) marker using VFP. The following C++ code snippet shows you how to decode the comment marker but assumes you are processing the marker stream.

unsigned int;
char comment[512];
int iIndex = 0;

length = inputstream.getBigEndianWord();
comment[iIndex] = inputstream.getByte();

for (iIndex = 1; comment[iIndex - 1] != '\000'
   && iIndex < sizeof(comment)
   && iIndex < length - sizeof(UBYTE2); ++iIndex)
   {
      comment[iIndex] = inputstream.getByte();
   }

I am in the process of testing an update (shortly to be uploaded) to my VFPImage extension fll that exposes the comments and application specific markers to the developer.

HTH
Neil