how can I perform interpolation between images

<<

<<

>>

>>

Hi, I have a set of 180 ultrasound images. I would like to apply linear interpolation between them to reduce the noise. Any help how I can do that? I appreciate ur help. Thank you, Yazan

Answer 1

[Image Analyst]

Interpolation? How would that reduce noise? Maybe you mean regression. But how many output images do you want from this set of 180 images?

[Yazan Awwad]

The 180 images contain an eraser that moved when the probe was moved. Using phase correlation, I found out how much exactly the movement was between each two consecutive frames. I shifted all the frames in respect to frame one,so now the eraser in all the frames is at same position. Now what i am trying to do is to get one big output image that only has the important features with reduced noise.

[Image Analyst]

Why not just average all 180 registered (aligned) images together?

[Yazan Awwad]

I did that, like I summed and normalized frame 1 and frame 2 and then the result i got i summed it and normalized with frame 3 and so on. But the result i got is what is only shown in the last frame (frame 180 ). Like one part of the frame was completely black (because when the frames were translated, what came in place of the translated area is a completely black area. So what i got as a result from sum and normalization is black area followed by the eraser(this is what frame 180 shows.

[Image Analyst]

Well obviously you did the sum incorrectly. Did you cast to double before you added them? Otherwise it will clip at 255, then when you divide by 180 you'll get 0. Or check Alex's suggestion - Prof Milanfar is a really bright guy with some clever algorithms that are very effective..

[Yazan Awwad]

I just did what u said, I casted to double and then did the sum and division by 180. It seems to be working but the output image is too blurry where in the original frames the shape of the eraser was brighter( I uploaded some samples of the frames and the output result). I was wondering if there is away to figure out what would be in the black area after translation (see Translated frame_180 with respect to frame_1.png as an example). please give me some suggestion how i could obtain a clearer output image of the 180 frames. I will take a look at what Alex's suggested. Thank you

Answer 2

[Alex Taylor]

This family of algorithms is known as super resolution. I suggest you review the literature. For example:

http://users.soe.ucsc.edu/~milanfar/publications/journal/SRfinal.pdf

[Yazan Awwad]

Thank you Alex, I will take a look at the paper.