Decent PDF compression products will segment scanned images into components and compress each according to content. This results in much smaller scanned pages which look identical to the original.
Examples include:
- Nuance Omnipage which uses MRC compression.
- Solid PDF Tools which uses CGM compression (my company's product).
Typically these kinds products will break the pages into parts that make sense from a compression point of view:
- photo-like images are downsampled (say from 300dpi to 150dpi) and then compressed with lossy compression (Jpeg or Jpeg2000)
- black text is compressed using lossless compression (CCITT Fax for example)
- "flat" color (like graphs, etc) is converted to palettized images.
- etc.