OneTaskBy Tim King · Updated
Compare Two Images Online
Find Differences Between Two Pictures and Get a Similarity Percentage
OneTask is free, with no sign-up, and built by one person.
If this tool saved you time, a coffee helps keep new ones coming.
How the Image Compare works
Compare two images online without uploading either of them. Drop in two versions of a picture, a screenshot before and after a change, two renders, two scans, and see exactly where they differ. Changed areas are painted red over a faded copy of the first image, and the stats give a similarity percentage, the number of separate changed regions and how different the overall tone is. Both files are read on your device and never uploaded.
How to compare two images online
The comparison is a straight pixel subtraction. The two images are lined up corner to corner, the second is resized to the first if their dimensions differ, and for every pixel the tool takes the absolute difference in brightness between the two. A pixel counts as changed when that difference is above the tolerance. At zero any difference at all is flagged, which will light up JPEG recompression noise and anti-aliasing across the whole frame; around 20 to 40 ignores that kind of noise while still catching a moved element, a colour change or a retouched area. Push it higher to see only the strongest changes.
Similarity is the share of pixels that fall within the tolerance, so 100 percent means the images are effectively the same and 90 percent means a tenth of the picture changed. That is a different question from whether two photos look alike to a person: a picture shifted by one pixel scores badly here even though it is visually the same. For a looser sense of resemblance the tone similarity figure compares the brightness histograms of the two images instead, which ignores position entirely and reports how alike their overall light and dark balance is.
Changed regions groups neighbouring changed pixels into blobs and counts them, which is usually the number you actually want when checking a UI screenshot or a proof: three changed regions means three things moved or were edited. The mean difference is the average pixel difference across the whole frame on a 0 to 255 scale and is a good single number to track between renders.
The three views serve different jobs. Highlight changes is for finding what moved. Difference map shows the raw subtraction as a greyscale image, black where the images agree and brighter the more they differ, which reveals subtle global shifts like a brightness or colour-balance change that the highlight view would either miss or flag entirely. Side by side simply places the two images next to each other at the same scale for a visual check. Images larger than 2000 pixels on the long side are scaled down before comparing.
FAQ
Why is similarity low when the images look identical?
One is probably shifted by a pixel or two, or one was resized. Pixel comparison is position-sensitive, so any offset flags large areas. Check the tone similarity figure, which ignores position, and the difference map, which shows an offset as ghosted edges.
What tolerance should I use for screenshots?
Zero for PNG screenshots from the same machine, since they should match exactly. For JPEGs or screenshots from different browsers, 20 to 40 hides compression and anti-aliasing noise.
Can I compare two PDFs or two web pages?
Not directly. Screenshot or render each to an image at the same size and compare those. Browser screenshots of the same page at the same window size line up well enough for the diff to show only what changed.
What if the images are different sizes?
The second is resized to match the first before comparing, which is fine when one is a scaled copy and meaningless when they are different crops. Crop them to the same area first.
See also: all Images tools, Audio, Video & Screen, PDF, SVG, Simulations & Screensavers, Generative Art, Chart Makers, 3D Model Viewers, Drawing Pads, Mockups & 3D Making, Diagram Makers, 3D Science, Tracker Music.
Related tools
Object Counter
Count coins, cells or particles in a photo by threshold.
Image Histogram Viewer
Plot the RGB and luminance histogram of a photo and spot clipping.
Image Resizer
Resize to a width, height or percent with the aspect ratio locked.
Edge Detector
Trace the outlines in a photo with Canny edge detection.
Blur-Up Image Placeholder
Make a tiny base64 LQIP and the code to blur-up images as they load.