Erroneous html cleaning in officeimporter when cleaning image tags having relative paths

Currently ImageFilter uses following code:

// OpenOffice 3.2 server generates relative image paths, extract image name.
int separator = src.lastIndexOf(File.separator);
if (-1 != separator) {
  src = src.substring(separator + 1);
}

Here the use of File.separator is wrong because url paths always use the character "/" which is platform independent. This code will fail to clean relative paths if executed on a windows environment.