In two stages to deal with etf feature

Clumping happens when trying to adjust parameters in reasonable range.

m1, m2 and stipplesize

m1 = 5, m2=3, stipplesize used
m1 = 3, m2=3, stipplesize used
m1 = 3, m2=3, use '2' to replace 'stipplesize'

Watercolor without ETF

They look similar to results with ETF feature. But results with ETF feature should be a little better.

Some examples for Watercolor

examples for watercolor

17000stipples with double size
same number, smaller size

same number, grayscale
10999stipples,f ewer stipples with bigger sizes

Some examples for Heightening

We can use sheep for ETF image

ETF feature with soft scaling function

To space out the ETF lines, I did it by dealing with the intensities under the mask. I showed the similar images to you yesterday. Now I remember why other pixels show evenly. It is because of the mask size.


A problematic image by ETF with not strong scaling function and try to reduce the stipples by shrink function for a dark imageETF feature

ETF with soft scaling function and try to reduce stipples for a lighter image

Evenly distributed stipples(190)

Hand-drawn stipplesFloating point paper

Our stippling

Fourier images for woman

All tests for woman image

Paper: 4000 stipples
Paper: 7000 stipplespaper: 14000 stipples

Our etfFeature-guided stippling

Measure based on CAH

woman: PSNR MSSIM CPSNR

CAhalftoning: 34.1692 ----- 0.0249 ----- 11.7397

14000:　 30.3356 ----- 0.0352 ----- 11.5094

7000: 48.9462 ----- 0.1291 ----- 12.53

4000: 57.0701 ----- 0.3082 ----- 13.625

our ETF: 28.1266 ----- 0.0807 ----- 12.0627

feature-guided:40.39 ---- 0.1346 ----- 15.2728

It seems the values cannot say a lot of things. But the MSSIM looks reasonable.

Fourier Spectrum