CODEWITHSUNDEEP
pythonpython-3.ximageopencvimage-processing
Sreevathsabr
Sreevathsabr

Reputation: 689

Comparing two Similar images which are been scaled in different way Or Compressed in Different way

I am Trying to compare two images which looks similar , but When I do pixel to pixel comparison ,they won't be similar . Here the Problem is the Image is compressed in different Ratios .

To give more Example , If we take the Facebook Profile picture and the Twitter Profile Picture and Compare Pixel by pixel , we will see images are NOT Same . But by look and feel, we will see Image are same (Image Width and Height are same)

I have tried one of the Library called ImageHashing which is available in Python , when I use Average hashing , even when I Have dark Line on image it will show as same , where Phash will have same problem as Pixel to pixel compare .

The Other way which, I thought was Template matching(OpenCV) , But I am not so convinced for Image Comparison

Is there any way can compare two similar images which are compressed or Sampled differently and get proper result ?

Upvotes: 0

Views: 1915

Answers (1)

user2448180
user2448180

Reputation: 46

You might consider the combined Mean Squared Error (MSE) and Structural Similarity Index (SSIM) processes taught in this tutorial:

https://www.pyimagesearch.com/2014/09/15/python-compare-two-images/

An MSE of 0 indicates a perfect match; A SIMM index of 1.00 indicates a perfect match. It's kind of arbitrary, but an MSE under 1000 and a SIMM Index above 0.5 would indicate a strong similarity despite differences in compression and angle.

Upvotes: 2

Related Questions