Although, with BLOB objects you can achieve this, it not the preferred way. You have to store the location of the images (similarly for other large files), when the query result turned, use the location to access the images.
Keep in mind, your images will be increased over time, so, in practice, we apply a directory hierarchy, usually two will be enough. One can use the first two letter of the name of the image, however, this will be unbalanced directory structure. The better idea is using hash based randomization to distribute the files evenly over the sub-directories. An example (in Java) can be found here.