Inspired by the Borel circle squaring problem — cutting up a circle into Borel pieces and translating them to make a square — I tried to draw a discrete version. I settled on a 169x169 square, and a pixelated circle* with radius 95.34, as the first decently large example to try. A simple search turned up this construction with 5 pieces.
2 pieces are easily seen to be impossible. It is probably not too hard to exhaustively prove that 3 pieces are impossible. So it leaves the question of whether 4 pieces can be achieved. Throwing this out in case someone smarter can figure out how to efficiently decide this with SAT technology!
* Consider a circle with radius 95.34 centered at the origin, we take the integer lattice points included in this circle and put a unit square at each lattice point.