I believe it's the math... I think the video feed is 720p. When you display 1080p, there's some, but generally un-noticeable distortion cause by the presentation compression, but when the same 4K image is reduced to the 720p, there's greater distortion. Just guessing.....
Like DJI-TIM states, the real video is on the SD or SSD....
What software are you using to view the picture? Is the software perhaps trying to adjust the pin cushion (wrongly) of the lens based on the data in the EXIF?