Hi Ahmed -
I confirmed this by adding a validation routine for new bootloader images that had version 1 of the header - the checksum was correct when done over (n-4) bytes - it also makes sense that the length (n) in version 0 headers are in 32-bit words compared with bytes in the version 1 header and therefore needs the x4 factor in the calculation only for version 0 headers.
The text "header checksum" is just incorrect - the header checksum is just the checksum of the header portion of the image - the image checksum is what is described.
Best Regards,
Paul