기울어진 스캔 페이지를 OpenCV로 자동 보정합니다.
클릭하여 파일 선택 또는 파일을 끌어다 놓으세요
PDF 파일 1개 이상
파일이 기기 밖으로 나가지 않습니다.
Pages with skew below this threshold won't be corrected.
Processing...
Select your scanned PDF with tilted pages
OpenCV analyzes and detects skew angles
Get your straightened PDF instantly
Each page is rendered to an image at the DPI you choose, then OpenCV analyzes it: it fits a minimum-area rectangle around the dark pixels and, as a fallback, measures the angle of dominant text lines with a Hough transform. The resulting angle is what the results panel reports for that page.
Pages whose detected tilt is below the threshold are left alone. The default of 0.5 degrees catches visible tilt without touching pages that are already straight; drop it to 0.1 degrees when you want the tightest alignment before OCR, or raise it to 1 or 2 degrees to correct only clearly crooked pages.
No. Every page in the output, including pages that didn't need correction, is written as an image at the processing DPI, so any existing text layer is lost. Run the result through OCR PDF afterward if you need searchable text; OCR works better on straightened pages anyway.
150 DPI is fast and accurate for ordinary letter-sized scans. Use 200 or 300 DPI for small print or fine detail, and 100 DPI when speed matters more than precision. The setting also controls the resolution of the output pages, so higher values produce larger files.
Output pages keep the original page dimensions. The rotated image is padded with white in the corners exposed by the rotation and then fitted onto the page, so content may shrink by a hair but nothing is pushed off the sheet.
It works best on pages with clear horizontal lines of text. Pages that are mostly photos, drawings, or diagonal design elements give the detector little to measure, so the angle can be off; check the results panel and keep your original for those.
After processing, a results panel lists every page with its detected angle and marks the ones that were rotated; pages under the threshold show their angle but are labeled as untouched. If you queued several PDFs, each downloads under its original name and the panel shows the last file processed.
No. The PDF rendering and the OpenCV analysis both run in your browser through WebAssembly. The engine is downloaded the first time you use it, but your documents never leave your computer.