Using the frame stacking technique, 200,000 frames all on top of each other produces this curious grid and step pattern (as well as some dust and wierd streaking):-
The steps around the lens opening are unexpected since they exceed the 8 x 8 pixel MCU. They appear to be two or three MCUs in size. We can only speculate that this is a ‘feature’ of the camera’s colour space transformation during JPEG encoding and interaction with the varying luminosity levels around the lens.
The pronounced grid is due to the nature of DCT compression. It is akin to a wire mesh through which the image is viewed. Quantization error and low-pass filtering of spatial frequencies a block at a time, magnifies the slight but cumulative edge discontinuity errors between MCUs. DCT behaves well at the edges of the data sample. This means that the DCT introduces the least edge-discontinuity between neighbouring pixel blocks. However, while the edge discontinuity is small, the slope of signal change at either side of a block boundary is nevertheless discontinuous. The discontinuity manifests itself as a fine virtual grid.
(Thanks to scottbb for the mesh causing DCT explanation .)