Ask HN: Best way (HW&SW) to digitize family photo albums? | Hacker News Reader

Ask HN: Best way (HW&SW) to digitize family photo albums?

4 pointspreya2k4 comments

Family has a bunch of photo albums laying around, that I'd like to digitize for better accessibility and durability.

It's mostly standard photos (like 10x15cm or 9x13cm) stuck on to thick white paper (using special "photo corner" tape, so they're removable) with handwritten captions below the photos. Photos are on the front and back of a page.

So, I need to scan whole pages. Are there scanners, that can do this in bulk for front and backside? I assume most top-feeding scanners will have problems with the thick paper and the glued-on photos. So do I need to do it page by page manually? What resolution and what file format should I use?

What about the software part? Is there a way to automatically cut the photos from the white background and save them to separate files?

Any experiences?

View on HN

> Family has a bunch of photo albums laying around, that I'd like to digitize for better accessibility and durability.

Don't do this for durability, unless you've already got a robust backup solution. Hard drives tend to fail predictably.

> I assume most top-feeding scanners will have problems with the thick paper and the glued-on photos.

The majority of page feeders (like in MFCs) have a tight 180° turn to flip the page down to the scanner plate. You may find a straight-through paper feeder, but also make sure thick media can be handled.

> So do I need to do it page by page manually?

Yup.

> What resolution and what file format should I use?

Continuous tone images (like portraits) are fine at 150-200dpi. High contrast line art is better at 300dpi, and depending on the printer, you can see up to 800dpi with your eyes (without a loupe). Standard line art film can be rendered at 1500dpi.

So, take the largest you want to print out at, multiply by ~300, then multiply by 1.5 (ish) to allow for cropping, and you know how many pixels you want. For example, for an 8×10, you want (8×300×1.5): 3600×4500.

You can save as TIFF to be lossless and allow for higher bit depths for more exposure adjustment headroom, but if you aren't mucking with levels, you really won't see JPEG artifacts, especially if you save at 90-95% quality. JPEG file sizes will be dramatically smaller than TIFF.

> Any experiences?

Like my Grandma told me, put dates on everything.

Consider adding a date (maybe with a label maker?) to each page before you scan it.

Flat bed scanner will give best quality. Table top DSLR camera will go fastest with a reasonable quality. For automatically extracting the photos from a digitalized page, I experimented with OpenCV Shape detection in a small custom Python code and had reasonable results. Can share my code if you are interested (it is really not much beyond what you find at OpenCV tutorials).

Seems that the setup to digitize with a DSLR is quite complicated (lighting, no reflections, etc.). I think flatbed scanner might be my best option. Thanks for suggesting OpenCV! Sounds like it's worth a try.

For a low-cost solution, you should consider Google PhotoScan.

https://www.google.com/photos/scan/

Reply on news.ycombinator.com