Image retrieval algorithm based on free-hand sketch

Abstract

The invention discloses an image retrieval algorithm based on a free-hand sketch. According to the method, a metric learning algorithm is used for obtaining a mapping function, the free-hand sketch and a real image are mapped to the same metric space, an L1 normal form in the metric space can directly represent the 'semantic similarity' of the sketch in an original space and a real photo, and the problem of cross-domain matching of the free-hand sketch and the real photo is well solved. Specifically speaking, a positive and negative sample pair is used for training the convolutional neural network, feature vectors of the photo and the free-hand sketch input in the space and labeled to be similar are put together, the feature vectors of images labeled to be not similar are pushed away, and an integrated retrieval method is put out finally. Experiments show that the retrieval performance of the method is greatly improved in the large-scale free-hand sketch retrieval data set.

Description

technical field [0001] The invention belongs to the technical field of information retrieval, and in particular relates to an image retrieval method based on a hand-drawn sketch. Background technique [0002] With the rapid development of computer technology and multimedia technology, a large number of digital images are produced. Effective image retrieval techniques are widely required in various fields such as remote sensing, fashion, publishing, agriculture, etc. How to quickly find a specific image in a massive image data set requires the use of image retrieval technology. Image retrieval is a technique to quickly find similar images in a massive image dataset through a given query. In recent years, example-based query has been widely concerned with the rapid development of information retrieval technology based on hand-drawn sketches. [0003] The key challenge of information retrieval based on hand-drawn sketches is to deal with the inherent ambiguity of the sketch ...

AI Technical Summary

Problems solved by technology

[0003] The key challenge of information retrieval based on hand-drawn sketches is to deal with the inherent ambiguity of the sketch itself, which is mainly manifested in the following three aspects: (1) Compared with the natural objects it depicts, hand-drawn sketches are more abstract in style, and the lines are not clear. rules, and the proportions of depicted objects are often different from the real ones

(2) When users draw sketches, they often draw according to the abstract concepts in their consciousness, without referring to real images, which leads to sketches with ever-changing appearance and structure, that is, hand-drawn sketches and real images are in different visual fields

(3) Due to the uneven domain knowledge and drawing skills of users, the sketches drawn have large intra-class changes, that is, different users will draw very different sketches for the same object, which has a great impact on the accuracy of retrieval

[0005] However, the main problem of traditional retrieval methods based on low-level feature description operators is that they are all based on an assumption that the visual domain gap between hand-drawn sketches and real images can be easily filled by hand-designed features.

However, the fact is that the conditions for the establishment of this assumption are often very harsh, requiring a lot of manual intervention to align and crop the hand-drawn sketches with the real image, and to perform preprocessing operations such as edge extraction on the real image, and the hand-designed features require Manually set a large number of parameters according to the characteristics of the images in the training dataset, and perform poorly on unknown images

Due to the inherent ambiguity of hand-drawn sketches, even after the above-mentioned various preprocessing operations, the edge map of real images is still difficult to match with hand-drawn sketches mainly composed of lines

Images