Wittgenstein never gave up the picture "theory" - that statements "picture" the facts, despite much being said that he did. The philosophical gulf between the Tractatus and the Philosophical Investigations is not as great as we might suppose.
Wittgenstein later amplified the Tractarian picture "theory" as "language games" in his Philosophical Investigations. The only distinction between them is that the picture theory pictured syntax while language games pictured syntax and semantics. The conclusion to be drawn is that semantics, that is, the many ways in which language is commonly employed, offers a much greater display of possible picture forms than syntax (logic) alone. This eventually made Wittgenstein give up the idea of a syntactical - logical - world endorsement, and the idea of the hegemony of a logically perfect language.
The reason why Wittgenstein was initially concerned only with the picturing of syntax is a historical artifact of academia. Russellian/Fregean logic, which Wittgenstein was introduced to and which he eagerly, youthfully and quickly, absorbed, and even logic today, is entirely syntactical. Logic did, and still does, looks after itself. Signs and symbols get shuffled about like beads on an abacus. Wittgenstein, through the influence of Sraffa, eventually saw through this historically forced, syntactically austere circumstance and began to realise that logic was not, after all, capable, in its "crystalline" syntactical purity, of picturing the entire world.
The picture "picture" (which is better than saying the picture "theory") cannot itself be made as a proposition of course because the statement that makes it or presents it itself employs a picture in making it. Rather, the picture is "shown", it is implicit in any coherent sentence whatever: Wittgenstein's Tractatus could have been about anything - the same point would have been made. But Wittgenstein is kind: he employs "elucidations" to invoke or suggest what he wants to indicate. But as Wittgenstein was committed to a logical, syntactically presented world he could not acknowledge the elucidatory form itself except as syntactical "nonsense" or as one of the species that "must be passed over in silence" (Tract. 7), which amounts to the same thing; though at the time he made no explicit link between them.
Thus we find Wittgenstein's proposal that his own propositions in the Tractatus were nonsense. Syntactically they were nonsense for the reasons I gave above. Equally, and non-syntactically, they were "that which must be passed over in silence", although the latter also refers to the non-syntactical framework of ethics. Here, even in the Tractatus there was a recognition that there were non-syntactical frameworks, though Wittgenstein placed these outside (passing over in silence) the framing structure of the "picture" because he was still commited to the syntactical, logical, world view.
Regarding Wittgensteinian pictures generally, Wittgensteinian Language Games didn't in the end, merge all pictures together in the nicely logical, syntactical way that typifies the presentation of syntactical elements or physical objects. The distinction between objects or the elements of syntax and the "elements" of pictures is that the former are commensurable and the latter are not. Most importantly, and a necessary condition for that, is that pictures are frameworks or necessary conditions for syntactical representation, and indeed for any object whatever (for example the framework of colour is not a colour but is the manifesting condition for any colour whatever). In this, Wittgenstein follows Kant.
Finally, as a historical addendum, the current inability of logicians and mathematicians to make much sense of Wittgensteinian logic and his distinction between syntax and the manifesting conditions for syntax, assists in piling error on error. Goedel, for example, unlike Wittgenstein and indeed Kant, failed to appreciate the logical distinction between syntactical and system or "picture" elements, a failure that culminated in the incompleteness theorems, theorems so often endorsed by the new transcendental realists (syntax is all) who make up the modern generation of logicians and mathematicians.