Image noise removal is part of image enhancement or image restoration?

Issue

I am working in digital image processing field, recently I am studying about Image Noise, I just want to know that whether this noise removal is part of image enhancement or image restoration.

I have read some papers on it, at some points authors are saying that image enhancement from a given noisy image, and on the other hand in some papers authors are saying that noise removal is image restoration process.

I got the following answer from dsp.stackexchange:

It’s task dependent: if you just want to make image “more beautiful” to print it, you would call noise removal “image enhancement”; if you want to make some kind of image processing and noise will strongly worsen resulting data, you will call it “image restoration”. Another way is restoration of broken and/or old photos: they have many corrupted zones. But in that case you can’t call corrupted zones noise because of their origin.

Please explain what is correct?

Solution

Well there is the wikipedia page http://en.wikipedia.org/wiki/Image_restoration

It says that

  • Image restoration is the operation of taking a corrupted/noisy image and estimating the clean original image

  • Image enhancement […] is designed to emphasize features of the image that make the image more pleasing to the observer

The difference between the two is, as explained in the dsp.stackexchange description you cite:

  • Noise removal for image restoration assumes that there is an underlying ‘ground truth’ image, and that your noise removal operation takes the image closer to the ground truth. As an example you can use de-noising on a photograph taken in bad lighting conditions to make the picture more similar to a better picture of the same scene.

  • Noise removal for image enhancement is not meant to approximate a ground truth image, but to highlight features (usually for display purposes). An example is the ‘airbrushing’ of photographs.

To complicate matters more, image restoration and image enhancement are not necessarily mutually exclusive (e.g., when the ground truth image is more ‘visually pleasing’ than the noisy image.

Answered By – alle_meije

Answer Checked By – Gilberto Lyons (AngularFixing Admin)

Leave a Reply

Your email address will not be published.