2
$\begingroup$

I was wondering if it could be possible to filter noise in an image using machine learning techniques. In my concrete case, I have important information as points (of the same size) that form a diagonal and a lot of noise in form of ""random"" points like a genome comparison (see figure 1).

For sure, there are ¿simpler? techniques to do that. I have implemented myself a simple heuristic to clean the image, for example, but It could be done using ML? maybe like a dataset with diagonals and other with random points? I usually use tensorflow so, if you know about something similar already implemented, it could be great to see it.

Genome comparison

$\endgroup$
2
$\begingroup$

Autoencoders can be used for Image Denoising (Xie) with exceptional results. Essentially, imagine a network where the input is your image (NxN) and the output is of the same dimension (NxN). If the hidden layer were the same or a greater dimension, the network would simply memorize your image. However, if you make the hidden layer smaller than your image, it must compress it. In this compression process, it removes noise. Not sure if it's good, but here is a tutorial in TF. Also see this video.

| cite | improve this answer | |
$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.