Mona Lisa Tile. After tiling for the first time I was… | by Tyler Altenhofen | Jun, 2021 | Medium
I started out by labeling each tile and taking a picture of them in groups. I used the checkerboard to find the homography for the floor plane. This allowed me to re-project the images of the tiles onto a flat plane and remove the distortion caused by the camera’s perspective. OpenCV makes this all pretty easy.
I sequentially read in each tile and place them over an image. I started out just placing each tile at the brightest location it could fit. This was accomplished by finding the max of a cross correlation of the tile profile run across the image.
The most apparent problem is that as we simply fill in the brightest sections of the image we loose any sense of proportional brightness. You can see below that for a simple gradient we just progressively fill from top to bottom.
To fix this we can blur each tile and subtract the created mosaic from the original image. This results in the program discouraging additional tiles in already filled areas and allows us to have gradations of luminance. You can see below that the simple gradient image now produces sections of varying tile density.
