How does browser perspective correction straighten angled photos?
Browser perspective correction maps the four corners of a tilted paper sheet onto a flat rectangle using a 3x3 homography matrix. By calculating the projective transform from your dragged quadrilateral to an upright grid, trapezoid distortion from holding a phone at an angle disappears cleanly.
Automatic corner detection and manual quad nudging
When you tap Use camera, the page requests your rear camera at up to 3840 x 2160 px via the MDN MediaDevices.getUserMedia() Reference with audio disabled. Each snapshot is scaled to 240 px on a temporary preview canvas to detect page borders automatically.
If a desk surface closely matches the paper color, corner detection falls back to a 4 percent inset frame. You can drag any yellow corner handle or focus a handle and press arrow keys to nudge coordinates by 1 percent (4 percent with Shift).
Mapping four corners with a 3x3 homography matrix
Inside the background worker, source photos are capped at 3000 px before computing a bilinear perspective warp up to 2400 px on the longest edge. Each destination pixel samples from the exact tilted quadrilateral in your original photo.
For flat flatbed or feeder scans that only lean by a couple of degrees without perspective keystoning, you can straighten skewed feeder scans directly after building a PDF.
Homography warping preserves straight table borders and crisp letterforms even when a photo was snapped at a 30-degree angle across a desk. Keeping the camera parallel to the paper still gives the sharpest text, while corner nudging cleans up any leftover skew.
Comparing Document, Black and white, and Colour photo looks
Choosing the right Look controls both shadow removal and JPEG compression quality. Document and Black and white divide each pixel by a local blurred paper-brightness estimate so overhead lamp shadows vanish into clean white paper.
| Look option | Shadow removal | Color output | JPEG quality | Best suited for |
|---|---|---|---|---|
| Document | Local brightness division | Full color with whitened paper | 0.85 |
Contracts, signed forms, and highlighted notes |
| Black and white | Local brightness division | Pure monochrome ink on white | 0.80 |
Text-only letters and minimum file size |
| Colour photo | None (preserves lighting) | Original camera colors | 0.85 |
IDs, book illustrations, and glossy brochures |
| Whole photo + Colour photo | Uncropped full frame | Original camera colors | 0.85 |
Whiteboards, posters, and square artwork |
As explained in the Library of Congress JPEG Format Description, JPEG quality factors balance edge crispness against file size. Switching from Colour photo to Black and white removes color chroma noise and drops JPEG quality from 0.85 to 0.80, shrinking multi-page scans dramatically.
If an older thermal receipt still looks washed out after scanning, a faint scan contrast booster can deepen light ink further.
Choosing A4, US Letter, or native page dimensions
Selecting Page size determines whether every warped sheet keeps its own proportions or scales onto uniform office paper. Matching standard dimensions avoids awkward scaling when recipients print multi-page packets.
Matching paper standards vs. preserving exact aspect ratios
Leaving Same as the page selected sizes each PDF page directly to your cropped rectangle with 0 margin. Choosing A4 or US Letter fits every straightened page onto uniform sheets regardless of whether one receipt was narrow and another was full-size.
When scanning expense slips for reimbursement, you can also extract line items into an expense report instead of filing raw page images.
Tip: Tap Whole photo below the corner editor whenever you want to include a full square photograph or wide whiteboard without cropping edges.
Building multi-page scans from camera and gallery photos
You can mix live camera captures and existing gallery files in a single PDF using the thumbnail strip below the editor. According to the MDN OffscreenCanvas API Documentation, rotating and filtering runs inside a Web Worker so multi-page batches stay responsive on mobile browsers.
Tapping Rotate turns the selected shot by 90 degrees clockwise without re-detecting corners. Selecting Delete page removes an accidental blur before you tap Make PDF.
Making scanned PDFs searchable and compact after export
Photographed pages start as flat raster images embedded inside PDF page streams. Adding text recognition or trimming redundant resolution turns bulky phone photos into fast, searchable office records.
Why camera scans start as image-only pages
A camera sensor captures pixels rather than character codes, so PDF viewers cannot highlight or copy words from a fresh scan. The exported file defaults to scan-YYYY-MM-DD.pdf with every page stored as an embedded JPEG compressed cleanly at 0.80 or 0.85 quality.
Adding a searchable text layer locally
To index words for Spotlight, Ctrl+F, or screen readers, open OCR PDF from the completion screen. It runs optical character recognition directly in your browser and overlays an invisible text layer aligned to the straightened page image.
Because high-resolution phone cameras produce up to 2400 px output sheets, running Black and white or compressing afterwards keeps multi-page packets easy to email. Every step stays confined to your device memory with 0 bytes uploaded to external servers.
If you need to re-order pages or drop a blurry capture later, the thumbnail strip lets you add or delete sheets before assembling the final file. Each exported PDF opens cleanly in standard desktop viewers and mobile previewers without proprietary plugins.
You can also switch between Document, Black and white, and Colour photo at any time before tapping Make PDF, letting you test contrast and file size on multi-page packets without re-shooting a single page.