Class SimilarImagesSceneRecognition<Image extends ImageBase<Image>,​TD extends TupleDesc<TD>>

java.lang.Object
boofcv.alg.similar.SimilarImagesSceneRecognition<Image,​TD>
All Implemented Interfaces:
LookUpSimilarImages, VerbosePrint

public class SimilarImagesSceneRecognition<Image extends ImageBase<Image>,​TD extends TupleDesc<TD>>
extends Object
implements LookUpSimilarImages, VerbosePrint
Identifies similar images using FeatureSceneRecognition. For each image which is added a set of image features (descriptions and pixel coordinate) is found using the detector. After all the images have been added fixate() is called and after sometime a fast but not 100% reliable model which be required for looking up which images are similar. In this case similar means that they are of the same physical scene but from a "similar" perspective. Then for each image all the "similar" images are found and the image features are associated. Theses associated image features are tested to see if the two images really are related using the logic specified by similarityTest. If they are related their pairwise information is saved for later fast retrieval. While building there is significant cost in building the retrieval system, this is in general much faster than considering every possible image pair and trying to match them. The main down side is that it will not be 100% reliable.