As LTjake suggests, you might want to create a subclass for each of the imaging libraries you support.

For example, you have a line() method, instead of defining that in Image::Mate with a huge conditional for each implementation, you simply define it in each of the derived classes. Then have Image::Mate::new() return an object of that type.

Imager does something similar for its font support - Imager::Font itself doesn't know how to draw text, it lets Imager::Font::(Truetype|Win32|FreeType2|Type1) specialize _draw() which does all the heavy lifting.

This is pretty basic OO.

Another issue I see is that you appear to load every one of the imaging libraries even though the user would only want to use one of them, and you then call import() in package main for that module.

The use of the c option to new() seems a bit inconsitent too - for GD this is used to decide between a paletted image and a true color image, for Imager it's used to decide between an 8-bits/sample image (24-bit color) and 16-bits/sample (48-bit color), but I suspect this is a case of trying to make sense of the different options the modules support.