How should I preform the feature extraction for a logo image?

I have 2 logo images, from them one logo is a fake logo and another is a real logo. The fake logo has a dark color backgroung but the real logo doesn't and now i need to do fearture extraction on these images but I do't know which mehod is best for this perpose. I need to extract a numberic value from these images.

Please someone guide me what should I do.

Real logos:

Fake logos:

[hamida]

Image Analyst, Sir do you have any idea about it?

Answer 1

[Aastha]

Hi @hamida,

As I understand, you are aiming to extract a numeric value from images of real and fake banknotes in order to classify them.

One approach is to crop the same region around the logo in both real and fake banknotes and use the "ssim" function in MATLAB to compute a similarity score between the banknote logo under consideration and that of a real banknote. This will yield a single numeric value representing the similarity to the reference logo. You may refer to the following MATLAB code snippet to do so:

[ssimval, ssimmap] = ssim(image_under_consideration, reference_image);

The "ssimmap" output is of the same size as the input image and contains a pixel-wise similarity, providing a more detailed comparison.

You may refer to the MathWorks documentation of "ssim" function for reference:

https://www.mathworks.com/help/images/ref/ssim.html

A second approach involves using CNN regression model to predict a numeric value for each image. For reference, kindly refer to the example lined below:

https://www.mathworks.com/help/deeplearning/ug/train-a-convolutional-neural-network-for-regression.html

Hope this is helpful!