How can fiber optic image conduits consist of only fiber core, no cladding? Looking at fiber optic image conduits, I found that they have a packing fraction (the ratio of the fiber core's area to the total area) of 100% (source here). 
In other words, they consist only of fiber core, no cladding.
But without cladding, wouldn't the light travel between the individual fibers, messing up the output image?
Or is the cladding so extremely thin that this is simply a result of rounding up to 100%?
 A: There is still a small layer of air between them. Moreover, they are probably single-mode fibers, so, even if they touch each other, it is very hard to provide energy transfer from one fiber to another.
A: It is probably a marketing 100% rather than strictly 1.000 - it's difficult to pack circles into a circle with a 100% packing factor.
It's to contrast them with older models made from larger individually clad and covered fibres which had a very obvious chicken-wire grid effect
A: Cf: ulexite and selenite minerals.  If the fused fibers are in direct contact - certainly after being bent during heating - they will leak light into each other.  At least two solutions:
1) Gradient index fibers.  The light never makes it to the surface. The fibers must still pack.  
http://www.pages.drexel.edu/~bel23/HCP.html 
Monodisperse circles hexagonally close pack to 0.9069% density. You lose a little more for fiber-fiber cladding spacing.
2) The fiber preform is a thick low index glass rod tightly fit inside a thin wall high refractive index glass tube.  One end is heated, fused, and drawn to create a remarkable length of thin cladded fiber with the same component size ratio.  These fibers may be bundled and drawn again to become much thinner.  Then bundle and fuse to form the rod. A huge number of exceptionally thin fibers are them suspended in an enveloping high refractive index matrix, never quite touching.  Look up the fabrication of channel plate amplifiers.
http://en.wikipedia.org/wiki/Total_internal_reflection#Derivation_of_evanescent_wave
In very rough terms, the fiber separation for 1/e leakage, 37%, is one wavelength (in the cladding).  Two wavelengths is 1/e^2 or 14%, three is 5%.  Green light is about 0.5 microns. For a cladding 1.5 refractive index, a 2-micron cladding separation is 1/e^[(2)(1.5)/(0.5)] or 0.25% leakage.  Cladding thickness can be thin versus fiber diameter.
The real world is dirty and has footnotes.
A: In this article:
Ultra Resolution Imaging with GRIN POF
(only first four pages are for free)
it can be read that there are two methods (as @Uncle Al writes)


*

*Normal SI fibers(Glass) can not be smaller than 5 µm because of cross talk and rapid diminished image brightness. The fibers has a cladding of no less than 1 µm (about two wavelengths of light). In these fibers light travels in straight lines.

*In GRIN(see article) fibers(Plastic) light travels in curved lines, and never reaches the surface. There are no cladding and the fibers can be drawn down to 2µm. This improves both brightness and spacial resolution. These are the conduits with a packing fraction(fiber to total area) of 100%.
In both image conduits, the fibers are hexagonal packed.
