We can think of every extended source as a collection of point sources. So, let us only think of point sources. When we draw ray diagrams we often say that an image of a point is formed where 2 rays meet. I was just as baffled as you are but then it clicked of what i think is corect.
Image forming means seeing. Our eyes have evolved to form images of point sources out of diverging rays. (This totally makes sense because you can only have divergent rays from a point source.) The mechanism of our eyes has cornea and lens which converge divergent rays.--so in short we have no problem focusing divergent rays onto our retina.
Here is the trick. If 2 rays are converging then they must be diverging on the other side.
So for a 2D scenario, the correct way to see an image is to put your eye behind the image such that the rays enter your eyes. These divergent rays will be converged onto your retina. Done
And for a 3D scenario, a light cone emerges out (for point sources again) so if you manage to get your head such that the cone enters your eyes you will see the image. But again for maximum clarity, put your head on the diverging part of the cone. :-)
Actually, a sharp image is formed if every point forms a point on retina and if every point falls on a region of retina instead of a point, a collection of such points-an extended source becomes blurred.
I cant be sure that my answer is correct as I didn't refer any source.. Require assessment from good physicists.
[![This is for the 2D scenario, apply the same to cone]]
The divergent rays will converge onto the retina by lens. Though only two rays have been shown many rays come out forming the complete image. FYI img is formed between 'C' and 'F'.