Prompt

How do I use OCR for low-quality phone photos of paperwork?

Artificial Intelligence / Computer Vision1 observationLast seen Oct 11, 2026

Latest observation

Oct 11, 2026OpenAI APIWeb search: off

To get good OCR from low-quality phone photos of paperwork, the biggest win is usually better image prep before OCR. Here’s a practical workflow.

1) Capture the photo in a way OCR can actually read

If you can retake the photo, do this first:

  • Use bright, even lighting; avoid shadows and glare.
  • Hold the phone parallel to the page as much as possible.
  • Fill the frame with the document, but don’t cut off edges.
  • Tap to focus on the text.
  • Use the phone’s document scan mode if available.
  • Avoid zooming digitally if possible; move closer instead.

2) Clean up the image before OCR

Low-quality photos often need preprocessing:

  • Crop to just the paper.
  • Deskew / perspective-correct the page so text lines are level.
  • Convert to grayscale.
  • Increase contrast.
  • Apply denoising if the image is grainy.
  • Try adaptive thresholding or binarization for faint text.

Tools that can help:

  • Adobe Scan, Microsoft Lens, Google Drive OCR
  • Tesseract OCR with OpenCV preprocessing
  • PaddleOCR or EasyOCR for tougher images

3) Pick the right OCR engine

Different engines behave differently on messy photos:

  • Tesseract: good for clean printed text, best when preprocessed.
  • PaddleOCR: often stronger on real-world phone images.
  • EasyOCR: easy to use, decent for mixed quality.
  • Cloud OCR like Google Vision, AWS Textract, or Azure OCR can be better for hard cases.

4) OCR settings that matter

If you use Tesseract, useful ideas:

  • Use the correct page segmentation mode:
    • --psm 6 for a block of text
    • --psm 11 for sparse text
  • Use the correct language pack if the document isn’t English.
  • If the document is a form, try OCR plus layout extraction rather than plain text only.

5) Post-process the text

OCR output from bad photos usually needs cleanup:

  • Spell-check obvious errors.
  • Use known context, like names, addresses, dates, and invoice formats.
  • For forms, validate with patterns:
    • Dates
    • Phone numbers
    • Email addresses
    • ZIP/postal codes
  • If needed, compare multiple OCR passes with different preprocessing.

6) A simple Python example

If you want a quick starting point with Tesseract:

import cv2
import pytesseract

img = cv2.imread("photo.jpg")

gray = cv2.cvtColor(img, cv2.COLOR_BGR2GRAY)
gray = cv2.bilateralFilter(gray, 9, 75, 75)
thresh = cv2.adaptiveThreshold(
    gray, 255, cv2.ADAPTIVE_THRESH_GAUSSIAN_C,
    cv2.THRESH_BINARY, 31, 11
)

text = pytesseract.image_to_string(thresh, config="--psm 6")
print(text)

If the page is tilted or warped, add:

  • perspective correction
  • rotation detection
  • line detection

7) Best practice for very bad photos

For really poor images:

  • Run multiple OCR passes with different preprocessing.
  • Try both grayscale and thresholded versions.
  • Use a stronger OCR engine like PaddleOCR or a cloud OCR service.
  • If possible, rescan from the original document instead of enhancing a bad photo.

8) What usually fails OCR

Common problems:

  • Motion blur
  • Reflections from glossy paper
  • Shadows over text
  • Curved pages
  • Very small text
  • Low resolution
  • Compression artifacts from messaging apps

If you want, I can also give you:

  1. a step-by-step Python pipeline for phone-photo OCR, or
  2. recommendations for the best OCR tool depending on whether your docs are forms, receipts, or typed pages.

Observation history

How did Obsurfable measure this prompt?

Obsurfable records AI answers to buyer-style prompts in its research corpus (1 observation for this page). Metrics are distributions over observations, not a single static ranking.

Which AI systems does Obsurfable collect answers from?

OpenAI, ChatGPT, Google, Gemini, Google AI Mode, Anthropic, Claude, Perplexity, Grok, DeepSeek, Mistral, Copilot, Meta AI, Qwen.

How can brands improve their AEO?

Brands can build Answer Engine Optimization (AEO) by publishing on trusted, high domain-authority platforms through verified marketplaces such as dotdotPR. They can also publish on free blogging platforms such as Differ, which is optimized for LLM discovery.

Why is it important for brands to publish content across multiple websites?

Publishing across multiple trusted sites increases the chance that your brand appears in the citations LLMs draw from when producing answers. Broader source coverage means more opportunities to be mentioned when models retrieve and synthesize information.

Want this interpreted for your brand?

Explorer is the free public corpus. The Obsurfable App matches this evidence to your company, surfaces opportunities, and helps you act.