How exactly to gauge the similarity between two pictures?

We have two team pictures for cat and dog. And every combined team have 2000 pictures for cat and dog correspondingly.

My objective is you will need to cluster the pictures by utilizing k-means.

Assume image1 is x , and image2 is y .Here we must assess the similarity between any two pictures. what’s the way that is common measure between two pictures?

1 Response 1

Well, there several therefore. lets go:

A – utilized in template matching:

Template Matching is linear and it is maybe perhaps maybe not invariant to rotation (really not really robust to it) however it is pretty robust and simple to sound including the people in photography taken with low lighting.

It is possible to implement these utilizing OpenCV Template Matching. Bellow there are mathematical equations determining a number of the similarity measures (adapted for comparing 2 equal sized pictures) utilized by cv2.matchTemplate:

1 – Sum Square Huge Difference

2 – Cross-Correlation

B – visual descriptors/feature detectors:

Numerous descriptors had been developed for pictures, their use that is main is register images/objects and look for them in other scenes. But, nevertheless they provide plenty of information on the image and had been utilized in student detection (A joint cascaded framework for simultaneous attention detection and attention state estimation) as well as seem it employed for lip reading (can not direct you to definitely it since I am maybe not yes it had been currently posted)

They detect points which can be thought to be features in images (appropriate points) the texture that is local of points if not their geometrical place to one another can be utilized as features.

You can easily discover more about any of it in Stanford’s Image Processing Classes (check handouts for classes 12,13 and 14, if you wish to keep research on Computer vision we recomend you check out the entire program and perhaps Rich Radke classes on Digital Image Processing and Computer Vision for Visual Results, there’s a great deal of information there which can be ideal for this hard working computer eyesight design you are attempting to simply take)

1 – SIFT and SURF:

They are Scale Invariant techniques, SURF is a speed-up and available type of SIFT, SIFT is proprietary.


They are binary descriptors and therefore are really fast (primarily on processors having a pop_count instruction) and certainly will be utilized in a way that is similar SIFT and SURF. Additionally, i have used BRIEF features as substitutes on template matching for Facial Landmark Detection with a high gain on rate with no loss on precision for both the IPD in addition to KIPD classifiers, although i did not publish any one of it yet (and also this is merely an incremental observation in the future articles thus I don’t believe there clearly was harm in sharing).

3 – Histogram of Oriented Gradients (HoG):

This can be rotation invariant and it is employed for face detection.

C – Convolutional Neural Networks:

I’m sure that you don’t would you like to utilized NN’s but i believe it really is reasonable to aim they truly are REALLY POWERFULL, training a CNN with Triplet Loss could be actually good for learning a feature that is representative for clustering (and classification).

Always check Wesley’s GitHub for a typical example of it’s energy in facial recognition making use of Triplet Loss to get features after which SVM to classify.

Additionally, if your trouble with Deep Learning is computational price, it is simple to find pre-trained levels with dogs and cats around.

D – check into previous work:

This dogs and cats battle happens to be happening for the number of years. you should check solutions on Kaggle Competitions (Forum and Kernels), there have been 2 on dogs and cats that one and therefore One

E – Famous Measures:

  • SSIM Structural similarity Index
  • L2 Norm ( Or Euclidean Distance)
  • Mahalanobis Distance

F – check up on other type of features

Dogs and cats is a simple to recognize by their ears and nose. size too but I experienced kitties as large as dogs.

so not really that safe to utilize size.

You could try segmenting the pictures into pets and history and then attempt to do area property analisys.

For those who have enough time, this guide right here: Feature Extraction & Image Processing for Computer Vision from Mark S. Nixon have much information about this type of procedure

You can look at Fisher Discriminant review and PCA to produce a mapping while the evaluate with Mahalanobis Distance or L2 Norm

Leave a Reply

Your email address will not be published. Required fields are marked *