How to correct an image perspective
Use this tool to rotate a tilted image or map a rectangular subject photographed at an angle back to a rectangle. Choose between Deskew (simple rotation) and Perspective (four corners), then export the corrected result as a PNG. Image loading, processing, and export happen in your browser with the Canvas API; the local script does not request an image upload.
Choose a correction mode
Select Deskew (simple rotation) when the image needs one overall rotation. Select Perspective (four corners) when a document, sign, screen, or other rectangular subject has trapezoidal distortion. This mode maps the four selected corners to a rectangle.
Correct perspective with four guides
- Drop an image into the tool or click browse files. The visible file types are JPG, PNG, WEBP, and BMP.
- Select Perspective (four corners).
- Drag the four guide handles onto the subject's corners. Use Grid overlay as a visual alignment reference.
- Optionally click Auto-detect edges (assist) to estimate a starting position from edge contrast. Check and adjust every guide before exporting.
- Click Reset to image corners if you want to restore the guides near the image corners and start again.
Straighten an image with Deskew
Select Deskew (simple rotation) and use Straighten angle from -45 to 45 degrees. You can also enter a value in Angle (deg), which accepts the same range. To estimate an angle from the image, place the dashed-line handles on an edge that should be straight, then click Use dashed line as angle.
Set output size and fill exposed areas
Choose a setting under Output canvas size:
- Auto-fit to guides uses the four-corner guide lengths for output dimensions in Perspective mode. In Deskew mode, it keeps the original image width and height.
- Preserve original image size keeps the source dimensions.
- Custom size lets you enter the output width and height.
When rotation or warping maps an output pixel outside the source image, use Fill exposed areas to choose the result:
- Transparent leaves the area transparent.
- White fills it with white.
- Edge-extend uses the nearest source edge pixels.
- Crop to valid area shrinks the output to a sampled area containing valid source content.
Export and file handling
Click Export corrected image to download a PNG. The filename uses the source name with -corrected.png appended. The page does not document a maximum file size, maximum pixel dimension, browser compatibility matrix, or performance limit.
Limits and what the tool does not do
This tool performs geometric correction only: rotation and perspective warping. It does not remove objects, read text, or reconstruct missing detail. Auto-detect is an assist, not a finished correction, so review the guides manually. Guide handles are kept within the source image bounds, although transformed output pixels can still fall outside the source and use the selected fill behavior.
Frequently asked questions
What is the difference between deskew and perspective mode?
Deskew applies a single rotation to straighten an image that is tilted but otherwise flat-on. Perspective mode lets you drag all four corners independently, correcting trapezoidal distortion from an off-angle shot of a rectangular object.
Does auto-detect find the corners for me automatically?
Auto-detect is an optional assist that estimates a starting position for the guides from edge contrast in the image. It doesn't read or understand the content, and it never applies a correction by itself - always check and adjust the guides before exporting.
Will this remove objects or fix blur in my photo?
No. This tool only performs geometric correction - rotation and perspective warping. It doesn't remove objects, read text, or reconstruct missing detail.
Are my images uploaded anywhere?
No. Loading, transforming, and exporting all happen locally in your browser using the Canvas API. Nothing is sent to a server.
What file format does the export use?
The corrected image downloads as PNG, which preserves transparency if you chose the Transparent fill option.