Bitmap Data Similarity Algorithm

Question

I would like to compare two bitmaps or parts of a bitmap and see how similar they are.

I came across some similarity algorithms for string data such as Levenshtein distance and Yaro-Winkler distance. they obviously don't help when it comes to raster data.

Can anyone suggest an algorithm to compare how the same bitmap images are?

---

EDIT

Thanks for the ideas, links, etc.

Although all the information is useful to familiarize yourself with this topic, I wonder how to create images, something more in accordance with this:

Answer 1

You can use Hausdorff image comparison , but keep in mind that this expects binary images.

It should be noted that visually similar images can have significantly different basic pixel representations. Human perception makes computer visual similarity quite complex. Hausdorff does a good job of this, allowing two different images to be praised when the overall similarity is the same.

If you want to calculate data distances, you can use any distance metric that you like and do it right in pixels. I especially love the Mahalanobis distance for such comparisons.

Answer 2

Here's an interesting link from a similar question . Using your own neural network training to recognize similar images.

Answer 3

You want to use the Hilbert curve to split the image and read the rgb values. You can then use the fast Fourier transform to write the image to a discrete analog signal. You can then save this in a database and compare it with other results. The results should be very good. In fact, this is similar to the Quantiziser step in JPEG compression (Hilbert curve), except that JPEG compression uses the Morton curve.

Answer 4

Neural networks also bring a solution. You can use the Kohonen network (it will group your images into classes and give a topological map) or use the Hopfield network (provide some representative images, and images after training will be classified on the network into images provided in the training mode).

Kohonen Networks: http://en.wikipedia.org/wiki/Self-organizing_map

Hopfield network http://en.wikipedia.org/wiki/Hopfield_network