As a rule images prepared for regular HTML documents have a resolution, which is not sufficient for print or for PDF. It is because print resolution typically ranges from 300dpi up to 1200dpi, while screen resolution is usually only 72dpi to 96dpi.
PD4ML never resamples images, but embeds them “as is” (JPEGs and some species of PNG) or re-compress them non-lossy way to generic PDF image formar (GIFs and the rest of PNGs).
A solution would be to use more detailed source images. For example:
You’ll get better results if you add a hi-res image (let’s say 1000x1000px) to your doc with an image tag defines smaller dimensions 100 width=100>. It will be embedded to PDF with the all detailization of the original 1000x1000px image (good zoomable / good printable), but shrinked to 100×100. That would allow you to keep the original HTML layout not broken by the huge 1000×1000 image.
Also we would recommend you to use JPEG image format (probably with the highest image quality settings) for such big images. JPEGs are embedded “as is” and it keeps the resulting PDFs compact. For example, GIF needs to be recompressed, which makes the resulting PDF to grow.