You want to detect some of the low-level structure in a PDF file, right? Perhaps the pagedump.pl script, which comes in the PDF CPAN package in the "examples" subdirectory would be of some help.

Here is the start of what it dumps about the first page of The Perl Journal:

% perl pagedump.pl 0301tpj.pdf 1
Page 1
    Dictionary
        <<
        Name: /CropBox => Array
            [
            Number: 0
            Number: 0
            Number: 558
            Number: 756
            ]
        Name: /MediaBox => Array
            [
            Number: 0
            Number: 0
            Number: 558
            Number: 756
            ]
        Name: /Rotate => Number: 0
        Other: Page_Object => Object: 402 0 R
        Other: Resource_Object => Object: 434 0 R
        >>
...
[download]

You can probably find a distinct set of components for your image-only cases.

Update: Mr. Muskrat and I seem to have different interpretations of your question. I read "detect that" to mean "detect that a file (which is already known to be a pdf file) contains only images rather than images plus text or text alone."