Wash Obser Obits The Hidden Stories Behind The Obituaries

Cleaning Faded, Stained Obituary Pages for Archive Work

Most people don't realize how bad newspaper obituaries degrade. I've spent years pulling yellowed pages from microfilm and digitizing them, and the standard scanner output is almost never usable for research or publishing. The ink fades, the paper stains, the typography gets crushed by print age. You need to process these images before anything you do with them will actually read correctly or look presentable. Here's the practical workflow for what I call Wash Obser Obits The Hidden Stories Behind The Obituaries — the image preprocessing pipeline that actually matters when you're working with degraded obituary material.

Getting the Raw Image Right

Start with the source. If you're scanning from physical copies, use at least 600 DPI. Newspaper text at that resolution gives you enough pixel data to work with during the wash stage. Going higher, like 1200 DPI, usually just bloats your files without adding usable detail for this kind of work. Set your scanner to produce a TIFF or an unoptimized PNG. Skip JPEG if you can help it — the compression artifacts compound when you start applying filters, and you lose text edge sharpness that matters later. Before you touch any software, flatten the page. Paper curls from old magazines and newspapers resist lying flat even with weights. I use a glass platen scanner when available. For larger documents I've worked with, I lay the page under a clean sheet of acrylic and let it rest for about thirty minutes before scanning. It reduces the wave distortion enough that the processing steps afterward behave predictably.

The Wash Processing Stage

This is where the actual work happens. "Washing" in this context means removing stains, tonal shifts, and background noise while preserving the text. I use a combination of GIMP and command-line ImageMagick for batch operations, though Photoshop works fine for single-page jobs. First, convert the image to grayscale if it isn't already. Color information from aged newsprint rarely adds useful data and only complicates the thresholding step. Then apply a curve adjustment to lift the midtones. Old newspaper backgrounds tend to sit in a muddy gray band. Dragging the midpoint up 15 to 20 percent usually separates the text from the paper noise without blowing out faint print. After the curves, run a non-destructive levels adjustment. Set your black point to around 10 and your white point to 240. This compresses the tonal range into something cleaner. Don't push these values too far or you'll lose light or faded characters entirely. I've seen people crank the white point to 255 and wonder why half the obituary text disappeared from their final result.

For stain removal — the yellow/brown discoloration that comes with age — I use a localized color correction layer. Sample a clean area of the paper if one exists, then use that as the white reference point. Apply it across the whole layer and the yellow cast drops significantly. If the page has no clean area, which happens often with water-damaged copies, set your white point manually using the levels histogram. Find the rightmost peak that represents the paper tone, not the text, and map that to pure white. The final processing step before export is a gentle unsharp mask. 50 percent amount, 2 pixel radius, threshold of 3. This sharpens the character edges without introducing halos around the text. Halos are a problem because they create false artifacts during OCR and make visual reading harder for anyone going through the collection later.

Common Pitfalls When Processing Obituaries

People tend to over-sharpen and over-contrast. The instinct is to make the text pop, but obituary columns from the 1940s through the 1970s often have very thin, lightweight typefaces that dissolve under aggressive processing. If your output looks like every character is surrounded by a white glow, you've pushed the threshold too hard. Dial it back. Another issue is uneven lighting from the original scan. I had a job last year working with a bundle of 1962 Cleveland Plain Dealer clippings where the photographer had clearly used a desk lamp setup instead of a proper document camera. One corner of each page was significantly darker than the rest. Standard global adjustments couldn't fix it without crushing the text in the bright areas. What worked was a local luminosity mask. I masked the dark corner, applied a separate curves adjustment just to that region, and blended it at about 40 percent opacity into the rest of the image. It took maybe twenty minutes for four pages that would have been impossible to read otherwise.

Beyond Basic Cleaning

Once your images are washed and cleaned, the next decision is how you organize and store them. I recommend keeping the original scan untouched in a separate folder. The processed version goes in its own directory. This way you can always go back and reprocess if you find a better method later, which you will. For OCR accuracy on processed obituaries, Tesseract 5 with a custom trained data file set for newspaper typefaces performs noticeably better than the default model. If you're processing more than fifty pages, training a basic model on a sample of that specific paper's typography cuts your manual correction time by roughly sixty percent. I use a simple command-line setup with jTessBoxEditor for font training, and it takes about forty-five minutes to get a usable box file for most American newspapers from the mid-twentieth century. One thing the process doesn't fix: handwritten annotations in the margins. Some obituaries have notes from family members added later in pencil or ink. Image processing will often make these illegible or remove them entirely. If margin notes matter for your research, scan those pages separately at a higher resolution before applying any wash filters, and keep those scans in their own folder.

The whole pipeline from raw scan to ready-to-publish image runs about fifteen to twenty minutes per page for a trained workflow. New setups will take longer — expect forty-five minutes to an hour while you're dialing in your settings for the first time. File sizes for a properly processed 600 DPI TIFF run between eight and fifteen megabytes depending on page density. Factor that into your storage planning if you're working with a large collection. This method works well for newspaper clippings, magazine obituaries, and most printed funeral program scans. It breaks down on items that were originally photocopied onto poor-quality paper, or on items with heavy creasing that shadows can't be corrected through software. For those, the only real fix is re-scanning with cross-polarized lighting, which is a different process entirely and requires equipment most people don't have access to.