What is Content-based image retrieval?
by Sergey Egorov
Wikipedia:
Ok. It is absolutely correct.
All web search engines leaders, such as Google, Yandex, Yahoo, Ask and etc., find multimedia content by means of text descriptions. Billions of images are tagged manually by keywords, so we have got used to search multimedia files by text queries.
But this approach have two cardinal and critical drawbacks:
- text descriptions are very subjective and therefore far from perfection, the same image can be depicted by absolutely opposite keywords;
- tagging multimedia files is very hard, time-taking and staff-resources-taking work.
And what is CBIR? It is a combination of different areas of knowledge, such as pattern recognition, object matching, machine learning, wavelet filtering and so on. CBIR is devoted to understanding visual characteristics of images without any text descriptions.
Try to think it over! The major part of information we perceive through our eyes, the major part of information we perceive at first several seconds, and even at fractions of a second. Most of us are lazy and don’t like to read texts, especially foreign texts. We like books and journals with illustrations, we prefer catalogues of goods with their photos, we choose the pleasure resort for our vacation looking at photos of hotels and sightseeings. You may continue this list infinitely! Oh, even small kids and children are able to distinguish basic colours and simple objects. They can’t read and write, but they perceive everything around them through their eyes.
Modern techniques and algorithms are able to understand images content even nowadays, it isn’t fantastics. It is reality and it is our nearest future!
Written by Sergey Egorov
No tags for this post.