This 'homunculus' view of the nature of perception, though it's related, shouldn't be taken as exclusively representing the views of
homunculism which is a view of the mind which holds that the apparent unity of the psyche is in fact an illusion, and that our mental functionality is accomplished in fair measure by a host of (to some extent competitive and conflicting) 'homunculi', understood as approximately '
software agents' each with specific triggers and goals, which each strive to actualise their intentions and points of view by bringing them into awareness, or the '
locus of control' of the subject.
In relation to the apparent problem of "what is it that is seeing the homunculus who's seeing the image", a straightforward commonsense view might suggest that, in fact, the infinite regress is not required. Unless you believe that esse est percipi (a rather unfashionable view: that to exist one must be seen to exist) there's no need to doubt the existence of an "inner viewer" on the grounds that it's not seen, and so we don't require an infinite regress of viewers each of which sees the ones "further up" and is seen by those "further down". It may be said that the 'inner viewer' is not seen: in the plain, ordinary sense in which we see stuff we are looking at, we don't see any 'inner viewer' at all; the whole notion that either we or something else does see this shadowy entity, in some strange, twisted, philosophical sense of see, seems to be confused and incoherent. To hypothesise the existence of an internal viewer, to think about one, is not to see one, and should be no more problematic than the consideration any other idea or concept
A more radical, Humean scepticism may be applied to the existence of the 'inner viewer' itself. We have excellent evidence for the existence of our perceptions - we're immersed in the things, and we should not really doubt the reality of what is, after all, our only means of accessing data from the "outside world". The existence of the subject who is having these perceptions, however, may be thought, perhaps, to be a grammatical fiction, induced by the convenience of the subject-object relation in language, and drummed into our heads from childhood by the fact that we're always called by the same name. As Hume pointed out, there's nothing that's literally constant in our perceptions. In that world, the Heraclitian doctrine that 'all is flux' was never truer. The 'inner viewer' or 'self' is much more in the nature of a hypothesis or a convenient manner of speech than a concrete reality, it might be thought.
There's much philosophical speculation about questions of personal identity - when all our atoms are (at least I believe) replaced after a seven year period, when our beliefs change, our knowledge grows and fades, habits are acquired and lost, it is open to question whether there is in fact a constant underlying reality to the shifting sands of our mental and physical lives which we are justified in using the word 'self' to describe. The view of homunculism is sympathetic to this scepticism, because even the homunculi (or "software agents") which constitute our mental toolbox and furniture may not be static or permanent - after all, over a lifetime we'll likely change our mental behaviour in relation to almost anything you can name. This may then be said to represent a 'homunculus problem' of an altogether bleaker and more radical kind, to which the only alternative seems to be some kind of 'essentialism' - viewing the self as a transcendent reality whose existence is not imputed by the constant modulations of its 'accidental' components.