Vectorize the images

Another comparison with two thresholds (128, 224)

Very close to their quality now and sometimes it looks better. Next to introduce smoothing.

Using two thresholds (192, 224) to fix the Lena and the sheep

Get more black back. Have to think about some measures for blackness and for the structure as well to obtain a good adjustment.

Using three thresholds (96, 128, 192)

Test the local histogram and find the maximum one. Decide the threshold based on the maximum one. Testing on the sticky results. Threshold should be as the adjustment to decide the blackness used. Structure quality is from the sticky mask but should introduce some adjustment also to control the structure details.

Using local average intensity as the threshold for each mask

It splitting the black. I am thinking if the gradual transition of the threshold value in the neighbouring area can improve the clustering.

Apply this to the results from the sticks filter

It means it preserves most of the contrast from the sticks filter. It also seems the thresholds should be adjusted. Shall we find an optimal adjustment?

Better, when reversing the contrast-aware distribution

The reason for this is because the priority scheme is working for that. They are still worse than traditional hard thresholding. I am guessing if adding the adaptive threshold control will be much better than this. 

Still black and white--adjust the threshold

It means at least the local adaptive threshold should be given a try later. 

T=128

T=180

T=200

Creating a priority map based on distance and adjusting the error based on a distance map from feature edges

The adjustment for the priorities based on a distance map cannot show a  lot of change. This is tested on texture edges.

 scaling (+)

 scaling( -)

The adjustment for the error is similar to previous results a lot.

 texture edges

 the etf edges 

Flipping the error sign to get black and white

Interesting, the black and white is created by only flipping the error sign when doing halftoning. Although it is not good for the structure, it shows the possibility and then we can think about the structure way.

Sparse version