You didn't mention the MPF-II's release: That was, apparently, 1982.
There were earlier Japanese Microcomputers than that, notably the NEC PC-8000 series (released in 1979) that used a Katakana/Kanji character set (that had to be removed/changed when preparing a release for the US market in 1981)
Even the Hitachi Basic Master (released 1978 and commonly considered the first "Japanese Microcomputer") used a Katakana character set.
If "desk size" falls into your microcomputer definition, the Toshiba TOSBAC might have been be the first with reasonable Kanji support.
A short comment on your question (too long to fit into a real comment): I think you are largely overestimating the complexity of the problem. Even entirely dumb telex machines could transmit and receive Kanji telegraphs right from the beginning - The operators simply used codebooks to transfer to and from a Japanese text into code combinations of the 32 possible characters in a telegraph (If you think about this, an English Dictionary that tells you how to build words from characters is not much different from such a codebook) . So, the only real problem is the method of input and output, rather than internal representation.
Character-based writing has a similar concept to Kanji - we classically call that words - And we had word processing software for roughly 30 years before computers were good enough to actually have a concept of words and offer input methods for them (with that I'm referring to real-time input checks whether something that's keyed in is actually an accepted word or not).
Input methods for Kanji were, in the beginning, rather a mechanical problem than a computational one - The first computers that allowed direct Kanji input worked with something closely resembling a codebook - A "book" of pages that were essentially keyboard overlays and depicted the glyphs, the computer detected what page you were in and could sort out the proper glyph from key and page (the TOSBAC mentioned above used such a method). Output: Well, as soon as the dot matrix printer was invented, that wasn't much of a problem anymore.
It is obvious that mechanical contraptions like the above are hard to put into a "Microcomputer" (whatever your definition of that would be), so the size of a TOSBAC is probably pretty close to the smallest that could be made before GUI input methods with contextual update and high-definition screens were available.