How to apply regression on images.

I want to do regression on one image taking other image as reference. I want to match these image DN values using regression.

[Walter Roberson]

"regression" requires a model. What model are you using? What are your predictor variables?

What is a "DN value" in this context?

[Image Analyst]

DN = Digital Number, basically, gray level.

[Tanya]

I have the residual values after applying 9 polynomial fir .. but how to display output image (which is regressed)

[Image Analyst]

I don't know what it means. Let's say that your reference image is cameraman.tif standard demo image, and your other images are cell.tif, moon.tif, and coins.png. Assuming they are all the same size, what do you want to do with those? Do you want to fit each and every pixel to a line?

If you simply want "to display output image (which is regressed)" you can use imshow().

[Star Strider]

Tanya — What is a ‘9 polynomial fir’? To me it implies a 9th-order finite impulse response filter. If it’s a 9th-degree polynomial you are fitting with polyfit, why?

[Tanya]

guess i m trying to normalize the image using other as reference. :(

Answer 1

[Elp]

How could be solved this problem ?

[Image Analyst]

It can't. Notice she never answered questions about what exactly that means. So, since no one has any idea of what she wants, we can't solve it for her. If you want to start a new question and explain it A LOT better than what Tanya did, you'll probably get an answer.

[Elp]

Thank you for your answer. My question is the following: i have a bunch of templates (let's say 180) and one input image. Like in a puzzle, i need to sum up some of the templates to recompose my input image. I was thinking to apply a linear regression ( non negative one, because i can not use negative weights), and after getting rid of the remaining weights that are associated with really high correlated templates. Are there a better way to do this?