Image Comparison techniques for same sized images

Hello, I have a reference image per say and a set of 100 images. I would like to go though those 100 images and pick out the ones that are the most similar to my reference image. When i say 'similarity' i am talking about visual and pixel similarities. All the images are same size and are background subtracted in the same way (same processing technique). My images look like something shown below. Different images show different development phases of the flow shown below. I have tried comparing histograms. Is there any better technique to this?

Answer 1

[Image Analyst]

Maybe the median absolute deviation or the mean absolute deviation

[P Pokhrel]

Thank you for your response. I have tried MSE approach before and i assume MAD will give me very similar results. For a stricter comparison, will cross correlation give me better results? I am relatively new to image processing using Matlab. I might be missing many easier ways to solve this problem.

[Image Analyst]

Cross correlation give you another image, usually with a big peak/spike at the center if the images are a lot alike. Why would that be better? I think a straight subtraction would be the "most strict", which is what the mean absolute difference is. Why do you think you need something different or better? A lot of image processing is new to me also (it's a big field), though I've been doing it a bit longer than you (35 years).