# Is it effective to use deep learning method to produce a 1D signal as output from a 2D image as input?

I have a 1D signal that will produce a 2D image after some image processing algorithm. Would it be possible and effective to use deep learning method to reproduce the 1D signal if I have the 2D image instead? If yes, what kind of deep learning neural network will be effective in this case?

In my application, I am looking at the noise images of the 2D synthetic aperture radar (SAR) images. Each image should look like random noise image (eg. tv static noise). I have the image processing algorithm that enables me to convert a 1D (or even a 2D) complex signal into that noise image. I am curious if it is possible for me to do the reverse process using deep learning instead because reversing the image processing algorithm is a complicated task. For dataset wise, I will randomly generate the 1D (or 2D) complex signals as the ground truths and the noise image as the training data. With the image processing algorithm, there should be no issue in generating the datasets.

Is there any relevant research / Github working on this method?

• Can you give more details about your specific problem? Why do you want to convert the image to the 1d signal? What task are you trying to solve? What have you found so far?
– nbro
Jan 4 at 10:33
• Thanks for the reply. I am trying to retrieve a 1D pulsed signal (radar) from a 2D synthetic aperture radar (SAR) image. Nothing has been done on my part yet. I am merely trying to establish a proof of concept. Jan 4 at 15:18
• I am not familiar with "2D synthetic aperture radar (SAR) images", but if there are patterns that can be used to convert the 2d images into 1d signals, it may be possible to do it with machine learning. I think it could help if you show us an image of "a 2D synthetic aperture radar (SAR)" and what you wish to obtain from that image, i.e. what exactly would be the 1d signal that you expect. Having said that, remember that ML is about learning from data. So, how much data do you have? Do you have labelled data?
– nbro
Jan 4 at 15:38
• In my application, I am looking at the noise images of the 2D synthetic aperture radar (SAR) images. Each image should look like random noise image (eg. tv static noise). I have the image processing algorithm that enables me to convert a 1D (or even a 2D) complex signal into that noise image. I am curious if it is possible for me to do the reverse process using deep learning instead because reversing the image processing algorithm is a complicated task. For dataset wise, I will randomly generate the 1D (or 2D) complex signals as the ground truths and the noise image as the training data. Jan 4 at 17:21
• I would recommend that you edit your post include these details because comments are temporary and are only meant to ask for clarification. In addition, after editing, your question will pop up as an active post and people will maybe read it again.
– nbro
Jan 5 at 10:07