You could, but it would be hard to get the 20fps you want.
For an initial test you could try just sending an array as json and see how it performs. Since it's all text the size of the data could vary a lot. Sending only the pixels that changed would be less data to send up to a point. Since you'd need to send color and position, you could actually need to send more data.
But you could go a simpler route, like just sending the player's input so both sides can recreate the same image.