Forging Art – A Tutorial

Centerfolds by Salavon

60s	70s	80s	90s

Title of Salavon’s work

Every Playboy Centerfold, The 1970s (normalized) 2002

Description

the photographs in this suite are the result of mean averaging every Playboy centerfold foldout for the four decades beginning Jan. 1960 through Dec. 1999

Playboy you say?

astonished woman wearing a hat holding a hand over hear mouth, high quality photograph, very high detail (DALL·E, inpainting)

Lenna

The Internet’s first poster girl, a standard test load used in the image processing community. The image was originally cropped from the November 1972 issue of Playboy Magazine […]

catb.org / jargon file

Get the shmutz

🧲 Playboy Centerfolds Ultra High Quality The Full (1953-2015) Year.torrent

Resolution : 3960х8640 - 4059х9060 | 8506x4080 - 9016x4098

Recreate original

arr = np.zeros((h, w, 3), np.float)
n_images = len(imlist)
for im in imlist:
    imarr = np.array(Image.open(im), dtype=np.float)
    arr += imarr / n_images

[…] result of mean averaging every Playboy centerfold foldout […]

Variations

Comparison to original

Comparison to original

Super Resolution

• Increase fidelity of image
• From lower resolution to higher resolution

SwinIR

• Works by overlapping enhanced patches
• State-of-the-art at the time
  • https://paperswithcode.com/task/image-super-resolution
• 4X resolution in each dimension
  • from 4059 × 9060 px
  • to 16384 × 36240 px
  • 190 × 85 cm @ ~500 dpi
• one model specifically trained on photographs
  \(^🫦^)/

More faithfulness to the original

Tried so far

• Unknown alignments
  • Tried 9 (3×3) variants
    • left, center, right
    • top, center, bottom
• Apparently different sources
  • Scan variants?

Image registration

Image registration is the process of transforming different sets of data into one coordinate system.

– Wikipedia

kevin-keraudren.blogspot.com

Use Simple Elastix or something like it!

High-res image

https://s.flx.ai/2022/digital-art-forgery/salavon-centerfolds-70s-forgery.jpg

Jupter Notebook

https://gist.github.com/flxai/497ce622ad6c28cc08da007fdd91ba6c

✌