When OCR output looks wrong — strange characters, broken words, letters swapped for numbers — the instinct is to blame the software. Usually it isn't the software. The recognition engine is doing its best with what you've given it, and what you've given it is an image that's harder to read than it looks to a human eye. Fixing the input is almost always faster and more effective than hunting for a better tool.
This is a practical guide to the image quality factors that actually move the needle on OCR accuracy, and what to do about each of them before you run a scan through any OCR tool.
Resolution
300 DPI is the floor for reliable OCR on typed text, not a recommendation for ideal conditions — it's the minimum. Below that, letterforms start losing definition at the pixel level, and the model begins guessing between visually similar characters: '8' and 'B', 'l' and '1', 'O' and '0'. These substitutions are the most common OCR errors and almost always trace back to insufficient resolution.
If you're scanning with a flatbed scanner, set it to 300 DPI minimum; 600 DPI if the document has small font sizes, fine print, or degraded paper. If you're using a phone camera, shoot from close enough that the text fills most of the frame — modern phone cameras have more than enough resolution, the problem is usually distance rather than sensor quality. Avoid digital zoom, which degrades sharpness without adding actual detail.
One format note: save scans as PNG or TIFF rather than JPEG where possible. JPEG compression introduces artefacts around high-contrast edges — exactly where letterforms meet white backgrounds — and these artefacts confuse OCR engines in ways that are invisible to the human eye but measurable in output accuracy.
Contrast
OCR models segment text from background by looking for areas of high contrast. Dark ink on white paper is the easiest possible input. Problems arise with anything that reduces that contrast: aged paper that's yellowed, coloured paper, light pencil marks, faded ink, or shadows falling across part of the page.
If your source document has contrast issues, adjusting brightness and contrast before running OCR is worth doing. Portimg's image editor lets you enhance contrast, fix uneven exposure, and remove shadows without any software installation. The goal is a scan where the text is as dark as possible and the background is as uniformly white as possible — even modest adjustments here can meaningfully improve accuracy on difficult documents.
Water stains, coffee rings, and background textures all compete with text for the model's attention. If a document has visible marks or coloured backgrounds, cleaning those up before OCR — cropping, whitening the background, removing obvious marks — pays off in cleaner output.
Alignment
Text that runs at an angle is harder for OCR engines to process because they typically analyse text in horizontal lines. A page scanned even five degrees off vertical can increase error rates noticeably, and pages photographed at an angle rather than from directly above are worse still.
When scanning physically, take a moment to align the document squarely in the scanner bed or directly below your phone camera. If you're photographing documents regularly, a simple document stand or book holder keeps the camera perpendicular to the page without you having to hold it. For documents that are already skewed, most image editors including Portimg's have rotation and straightening tools — fixing alignment before OCR rather than correcting the text output afterward is almost always less effort.
Lighting
Uneven lighting is the most common problem with phone-photographed documents, and the hardest to see when you're taking the photo. A shadow from your hand, a bright window behind you causing the camera to underexpose the page, or an overhead light creating a hotspot in the centre of the image — all of these create areas of the page where contrast drops and recognition suffers.
Flat, even light is what you're after. Natural light from a window to the side of the document (not behind you or the page) works well. If you're indoors without good natural light, two light sources on either side of the document eliminate most shadows. Turn off flash — it creates a bright hotspot in the centre and deep shadows at the edges, which is about the worst possible lighting pattern for document scanning.
Putting it together
The checklist before running any document through OCR is short: sufficient resolution, high contrast, straight alignment, even lighting. Getting all four right doesn't require specialist equipment — a clean phone camera, decent ambient light, and a flat surface to lay the document on is enough for most everyday scanning tasks.
For documents that need more work before they're ready — contrast adjustment, background cleaning, rotation — Portimg's editing tools handle that in the browser without any software to install. Once the image is in good shape, running it through OCR takes seconds, and the output quality difference between a well-prepared and poorly-prepared scan is substantial enough to be worth the extra two minutes of preparation.