Pattern Recognition Development Project-approximate degree calculation between images

I used to calculate the similarity between two face images when I was doing an illusion face. After thinking for a long time, I didn't come up with a solution. I forgot to send a help post on any forum, N after a long time, no one returned, and the problem was solved by his own ideas. Although no one has helped solve the problem, many children's shoes have suffered the same hardship. Now I have put my solution here for reference by the children's shoes who need it:

The similarity between the two images is simply the distance between the features of the images that can be described. Different types of images may have different types of effective features that can be described, some images have good gray information, while others require complex textures, colors, various histograms, and so on. However, they all share a certain degree of commonality, currently, I only use the features I have used as a reference. Other features can be similar to the following:

1. direct comparison of pixel gray scale of the image. Calculate the absolute values of all pixel gray difference of the two images and P = sum (| gray1-gray2 |), or the Euclidean distance;

2. Calculate the gradient image of the two images, and then calculate the absolute values of the differences between all pixels of the two gradient images and P = sum (| grad1-grad2 | ), or Euclidean distance;

3. Compare in the feature space, use the same group of samples as the training data to construct a certain space, and then calculate the projection vectors of the two images in the space, reserve the same length of the two projection vectors and calculate the Euclidean distance between the two projection vectors.

I hope this article will help you!

Http://blog.sina.com.cn/s/blog_672c5a470100qtht.html

Pattern Recognition Development Project-approximate degree calculation between images