Subject:
|
Re: Grayscale Mosaics
|
Newsgroups:
|
lugnet.build.mosaic
|
Date:
|
Sat, 8 Jan 2005 18:03:39 GMT
|
Viewed:
|
6270 times
|
| |
|
|
In lugnet.build.mosaic, Corey Sanders wrote:
| |
Maarten,
My mosaics are more than just a dithering of the 5 grayscale lego colors. In
MLCAD, using a dithering of the 5 colors, I create a palette of 9 colors that
I feel tricks the eye into seeing more shades, and therefore a clearer image.
Here’s a poor example:
On the left is the image with just the 5 color dithering. On the right is
the same image using my system. IMO it looks much smoother, especially in
person.
Corey
|
Maybe -first of all- we need a good definition of “dithering”:
Your left image shows no dithering at all. I see a similar conversion method as
the one I used on the mosaic of
my children.
I simply
divide the total image in X fields of darkness (in my mosaic X=11, in LEGO
mosaics X=5).
Example:
0% (black) < colour1 < 20%
20% < colour2 < 40%
40% < colour3 < 60%
60% < colour4 < 80%
80% < colour5 < 100% (white)
Altering these values may result in better images. I used this method because I
don’t have many 1x1 elements at my disposal in Ministeck: 1x1 elements represent
only 5% of the total amount of elements; therefor I try to avoid the use of them
wherever I can. (I even have an extra module in my program to remove as many of
the remaining 1x1 elements as possible: I only keep 1x1 elements if the contrast
to the adjacent pixels is at least two shades of gray.)
In my
Level 42 mosaic I used
the Floyd-Steinberg dithering algorythm: the error that comes with every
converted pixel is diffused to all its neighbouring pixels (“Error Diffusion
Dithering”). I have also tried the algorythms of “Burkes” and “Stucki”, they
are just a variation of the Floyd-Steinberg algorythm with almost similar
results. (I chose this method because it was the easiest for me to program, I
tried to implement the Hilbert-curve but that didn’t work out well...)
I think
I understand your method:
Instead of 5 colours, you use 9 shades of gray. For
the inbetween colours (1/2, 2/3, 3/4 and 4/5) you defined a checkerboard of 2
consecutive colours (actually you are still not using error diffusion
dithering). You obtain much better results with these 9 colours instead of 5,
but the main disadvantage of this method is that you will always need more 1x1
elements (the more 1x1 elements, the weaker your finished mosaic will be). The
Floyd-Steinberg dithering algorythm also results in many 1x1 elements, but never
in a checkerboard-pattern (1xN elements every now and then give the mosaic a
basic strength).
It is not a question of which is the best conversion method;
I choose a method in function of the materials that are available (and it is of
course also just a matter of taste).
| |
Your image would look like this:
|
I may have given a bad example: the original images are 750x640 pixels, I scaled
them down to 200x170 with great loss of resolution as a result. The original
images look at least as good as your conversion. Anyway, I don’t need this one
in grayscale: I already have this one in colour ;-).
Maarten
|
|
Message has 1 Reply:
 | | Re: Grayscale Mosaics
|
| (...) I guess we do need a good definition. :) In this image: (URL) The far left has NO dithering, the middle has the FS dithering (many programs do this), and the right is mine. (...) (snip) I use several different techniques but that is the heart (...) (20 years ago, 9-Jan-05, to lugnet.build.mosaic, FTX)
|
Message is in Reply To:
| | Re: Grayscale Mosaics
|
| Maarten, My mosaics are more than just a dithering of the 5 grayscale lego colors. In MLCAD, using a dithering of the 5 colors, I create a palette of 9 colors that I feel tricks the eye into seeing more shades, and therefore a clearer image. Here's (...) (20 years ago, 8-Jan-05, to lugnet.build.mosaic, FTX)
|
10 Messages in This Thread:
 
 
- Entire Thread on One Page:
- Nested:
All | Brief | Compact | Dots
Linear:
All | Brief | Compact
This Message and its Replies on One Page:
- Nested:
All | Brief | Compact | Dots
Linear:
All | Brief | Compact
|
|
|
|
