the process should look something like this:

on sending device side

1. aquire image
2. resize and convert to jpg
3. encode with base64 and publish via mqtt to server

on receive m5stickc side
4. subscribe data from mqtt server
5. decode image with base64 and save to variable
6. display image from variable on screen

I have a problem with 6 point

I skip points 1-4 that I won't have a problem with and I will deal with them later.
What I was able to get:
decode the received text as jpg and save it to a variable and to a file. And read it from the file and display it. Very frequent saving to flash memory will burden it tremendously and will shorten its life, so I want to skip writing. Unfortunately, I can't find any function to skip saving to the file and display the image directly from the variable. I was looking for a solution to my problem and the only thing I found was an analogous library for arduino where there is an identical function as in the micropyton and the one I am looking for, but on my hardware the compilation of code in arduino takes forever, which greatly extends the coding / programming / debugging process. I found a function / class in the micropyton
tft.pixel(x, y [,color])
but then I would have to send images as individual bitmap points and practically draw pixel by pixel but it would complicate my software for points 1-3 and extend the drawing time on the screen

working main part of the code:

text = '/9j/4AAQSkZJRgABAQAAAQABAAD ... ( ~ 3-5 kB encoded jpg file) ... '
temp = binascii.a2b_base64(text) # decode base64 ascii stream to binary jpg stream
f = open("res/temp.jpg", "wb")
f.write(temp) # save stream to file
f.close()
image0 = M5Img(0, 0, "res/temp.jpg", False) # create image object and load saved stream as jpg file
image0.show()