Answer by pklab for I want to develop a database of face image detected by camera. When camera detect a face, system will ask user to key in the person identity and save it in the database. Which means, my database consists of an image and ID. Or maybe it can contain a directory to where an image is save. It is an online enrolment. The data later on will be proceed for other purpose. So, I need a flexible database where I can update the data and call it back. But the thing is I don't know which medium is proper for me to use. I'd read about SQLite and CSV. But maybe there is other medium that can suits better.Any idea on how to achieve this?

Doesn't look and OpenCV question... agree ? Anyway: - DBMS is a way. SQLite works very good also for large record count and it can easily enclosed in your C++ application. Binary blob is supported too but it's up to you to store the images into the DB or use img files. - You could use [cv::FileStorage](http://docs.opencv.org/modules/core/doc/xml_yaml_persistence.html) but you can't have DB like query funcs. Some hard coded way (not really flexible) - You could save some meta data directly into the image file, EXIF for example, but this is [not supported directly by OpenCV](https://github.com/Itseez/opencv/issues/4344). Queries will be hard to do - At the end you could save the meta data casting your record as uchar and writing bytes at the bottom of your image (may be adding some rows) but this will be **break** if you save the image **with lossy** file format like .jpg