Method for searching visually similar images in collection of reference images such as road image, involves determining similarity between two images corresponding to formation of N-grams or gapped N-grams

摘要

The method involves extracting the visual features within the image. The features are quantized into visual words. The N-grams or gapped N-grams of visual words are formed based on the geometric arrangement of visual words in the image or scene. The use of an inverted file structure around the images is identified. The similarity between two images is determined corresponding to the formation of N-grams or gapped N-grams.