Tool
Perspective Correction for Photographed Drawings
Photograph a drawing by hand and the paper arrives as a trapezoid. Mark its four corners and this lays the sheet flat again, in your browser, without the photo going anywhere. It also shows you what the camera did, which is the part worth keeping.
What the camera did to your drawing
A camera records how big something looks, not how big it is. When you lean over a sketchbook, the near edge of the paper is closer to the lens than the far edge, so it images larger. Two edges that are parallel on your desk stop being parallel in the photograph: they converge. Follow them far enough and they cross. That crossing is the vanishing point, and it is the whole of perspective in one idea.
Switch on Show where the edges would meet above and the tool extends the edges of your own sheet until they cross. Most people have read about vanishing points. Rather fewer have seen one belonging to a piece of paper on their own desk.
Why it matters more than it looks
The error is not spread evenly. It is small near the camera and grows towards the far edge, which is exactly the pattern that makes an artist doubt their own eye: the drawing looks accurate at the bottom and increasingly wrong at the top. Measure that photograph against your reference and the tool will faithfully report drift that the camera invented.
A grid built from an uncorrected photograph inherits the same problem. The cells are not equal squares on the paper any more, they are squares in the photograph, and every one you copy carries a share of the error.
How the correction works
Marking the four corners tells the tool which shape was meant to be a rectangle. That is enough to work out the angle you shot from and undo it. The proportions are read from the perspective itself rather than from the trapezoid, so the sheet comes back not merely straight but the right shape: a square drawing stays square, and A4 stays A4.
The reading in percent is how far off square the photograph is. Below about three percent, correcting changes little and the tool says so. Above ten, measurements taken from that photograph will mislead you.
Getting a photograph worth correcting
- Fill the frame with the sheet, but keep all four corners inside it.
- Even light beats bright light. A window to one side and no lamp behind you avoids the shadow of your own head.
- Stand over the middle rather than at a corner. The correction copes with either, but a smaller angle keeps more detail.
- A sheet that contrasts with the surface lets the tool find the corners for you. White paper on a white table is the hard case.
Frequently asked questions
- Does my photo get uploaded?
- No. The correction runs on a canvas in your own browser, which is why it works with the window offline once the page has loaded. Nothing about the picture reaches us.
- What if the tool cannot find my sheet?
- It looks for the brightest region and declines rather than guessing, so white paper on a white table defeats it. Drag the four markers onto the corners yourself; the result is identical.
- Will correcting lose detail?
- A little, where the photograph was stretched furthest, because those pixels have to be spread over more area. The corrected sheet is sized to keep roughly the pixel count you started with, so nothing is invented and little is thrown away.
- Can I straighten a page from a book or a canvas on an easel?
- Anything flat and rectangular works: a page, a canvas, a print, a whiteboard. The method needs a shape you know was a rectangle, not a drawing of any particular kind.
- Should I correct before or after gridding?
- Before. The grid is measured against the picture, so a grid drawn on a crooked photograph inherits the crookedness. Straighten first, then send the result into the grid maker.