팀에서 검색엔진때문에 자료를 찾고있다가 구글 이미지 검색엔진이 궁금해서 자료를 찾아봤다.
My guess is that Google probably uses some combination of the following techniques and ranks the results using a proprietary algorithm.
Feature Detection (Image fingerprinting to look for exact match) & Search by Color & Visual Similarity Search
There are three commonly used feature detection algorithms for matching image deformation such as blur, rotation, scale, and illumination change.
They are SIFT, PCA-SIFT and SURF
- SIFT is slow and not good at illumination changes, while it is invariant to rotation, scale changes and affine transformations.
- SURF is fast and has good performance as the same as SIFT, but it is not stable to rotation and illumination changes.
- PCA-SIFT is the best but it has problems with image blur.
There is no perfect algorithm, so the choice depends mainly on the application, and what kind of trade-offs the application can tolerate.
Example: PixID http://ideeinc.com/products/pixid/
더보기 : https://www.quora.com/What-is-the-algorithm-used-by-Google-Search-by- Image-1