Smurf's Sky

Another comparison

2012-02-17T14:18:00.000-05:00

Both images show non-thresholded pixels in white. The rest pixels are shown as the original image.

(a) < T

(b) >T

Two images to separate the part above T and the part below T

2012-02-17T08:19:00.001-05:00

It is using adaptive threshold and dilates the sign-flipped error. Error is - I_cur when it is above T; otherwise, error is 255-I_cur. It did not use the CAH order and without ETF promotion.

(1) It shows white when I_cur is above T and the rest shows as (255-I, 255-I,0).

(2) It shows black when I is below T and the rest shows as (255-I, 255-I,0).

Raster image for low black

2012-02-15T12:03:00.000-05:00

black control

2012-02-15T12:01:00.001-05:00

Low black with promoting ETF priorities

High black without promoting ETF priorities

Some small modifications

2012-02-15T05:21:00.001-05:00

It is not using CAH order and the error is still using the older CAH values. The thing changed here is the weights. It is good for structure. Still I did not get the large black areas.

Results from adaptive threshold

2012-02-10T09:52:00.002-05:00

BW using Adaptive threshold with the CAH order

2012-02-10T09:47:00.001-05:00

Without error diffusion

2012-02-01T10:24:00.000-05:00

Without error diffusion

If without error diffusion, it is a thresholding method. The new modification is very close to this quality, but clustering pixels better. It means the error diffusion works but it is not the major factor to the structure quality now.

Vectorize the images

2012-01-31T13:33:00.002-05:00

Another comparison with two thresholds (128, 224)

2012-01-24T12:27:00.000-05:00

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

2012-01-24T11:10:00.002-05:00

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)

2012-01-24T11:01:00.001-05:00

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

2012-01-22T05:38:00.000-05:00

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

2012-01-22T05:31:00.001-05:00

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

2012-01-20T07:26:00.003-05:00

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

2012-01-20T07:24:00.000-05:00

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

2012-01-09T14:07:00.003-05:00

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

2012-01-09T11:17:00.000-05:00

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

2012-01-09T10:16:00.000-05:00

horse example to show the tone

2012-01-05T16:05:00.001-05:00

A little improvement

2012-01-05T15:45:00.001-05:00

Images from 3D models

2012-01-05T15:42:00.002-05:00

Very close to the scratchboard effects now

2012-01-05T15:24:00.001-05:00

X and Y (positive and negative)

2011-12-21T10:47:00.000-05:00

flat mountain

2011-12-13T12:42:00.001-05:00

Exclusion improvement

2011-12-13T07:08:00.001-05:00

both sides

One side

Old (below)

Hard specification

2011-12-06T14:14:00.002-05:00

It seems working now.

Not much improvement

2011-11-29T12:06:00.001-05:00

Very strict attractors

2011-11-22T12:23:00.002-05:00

Mountain Type I

2011-11-22T12:13:00.001-05:00

The configuration for the attractors still has a little problem. It should be fixed up later. Then those circular curves will look better.

A few more last tries

2011-11-15T14:33:00.002-05:00

Partly working

2011-11-03T15:29:00.001-04:00

Not exactly what we were expected. Scalar creates a lot of circles.
********************************Smoother tree**********************************

***************************** New way (more smoothed)*****************************

***************************** New way (less smoothed)*****************************

******************************Old way*******************************

Trails and trees

2011-11-01T16:04:00.000-04:00

Still CLD

2011-10-25T12:54:00.006-04:00

Those images are created before finding the vector graphics. A question is still there. Do we want the doubly edges?

A result from dilation

Thickness indication

2011-10-21T07:34:00.002-04:00

Recover some old CLD stuffs (not that high quality)

2011-10-21T07:33:00.003-04:00

Finally get those bugs gone. But the basic ides have a lot of limitations. Not sure if it is worth to recover them now. Should put more time on tree variations to change some artefacts and create more styles.

v: bw

2011-10-17T07:15:00.000-04:00

Different maps between H and V

2011-10-17T06:55:00.000-04:00

Canny test

2011-10-05T12:03:00.000-04:00

ETF and BW result (Example mentioned)

2011-10-05T11:59:00.004-04:00

H-way and ETF

2011-10-05T11:58:00.001-04:00

Get the basic GLSL working

2011-10-04T07:35:00.004-04:00

Next: testing illuminated line revisited

POV-ray renderer

2011-09-12T10:49:00.004-04:00

The next step is to implement Chen's rendering method used in her thesis.

Using two textures to replace the lines

2011-09-09T12:47:00.002-04:00

The alpha map should be added to remove the artefacts.

Triangulation based on edges

2011-09-09T07:25:00.000-04:00

Extract edges from three methods: canny, etf, and thresholding gradients. Then construct a triangulation from the points on the edges.

Gradients might be a better way for the irregular embroidery

2011-09-07T10:22:00.008-04:00

Choose a threshold of gradient magnitudes and densely draw short straight lines for each pixels. Since we do not care how many stitches, gradients decide the boundaries nicely. Then maybe we can think about the inner directions to have some contrast.Have to find a way to distinguish three contents: background, texture, and smooth parts.I am thinking to interactively define them if we cannot find an automatic way in a short time.

A few more examples by the over-smoothed field

2011-09-06T15:04:00.002-04:00

Smoothness of the field

2011-09-06T14:59:00.001-04:00

Randomly place 4000 stipples and use different smoothing times to draw a short line for each stipple. The line direction is based on the vector field.

Replace a stipple with an 'x' pattern

2011-09-06T10:30:00.000-04:00

Rendering parts should be added for effects.

Original intensity + gradient map

2011-07-06T13:33:00.002-04:00

Increased intensity (1.5 times) + ETF

2011-07-06T13:31:00.003-04:00

Mainly based on the quality of the ETF edges.But it looks better. It might go to the Minimal artistic rendering paper since theirs is based on tone balance.

Black and white (original intensity + ETF)

2011-07-06T13:30:00.002-04:00

Only based on the error to determine the color (black or white) in a raster scanning order. The problem is because in the dark area the etf edges should be white. But right now it is not. And another factor is that the error is not diffused now.

Radiosity Testing for Cornell Box (8 iterations)

2011-05-17T12:39:00.002-04:00

Fibers

2011-05-17T12:24:00.001-04:00

Log() applied for viewing

2011-05-17T12:23:00.002-04:00

Lights on different Distributions and diffuse uniformly

2011-05-17T12:22:00.000-04:00

All after 2 iterations.

randomly place 40 dots

Randomly place 2000 dots

It's a smoothing effect.

Another Lena

2011-04-28T03:15:00.004-04:00

Artificial Mosaics

1007 tiles

993 tiles

832 tiles

666 tiles

Small number of tiles

2011-04-27T15:41:00.001-04:00

590 tiles

1300 tiles

518 tiles

772

Lena (tessellation)

2011-04-27T14:27:00.002-04:00

Basic results(III)

2011-04-11T13:55:00.009-04:00

Basic results(IV)

2011-04-11T13:34:00.007-04:00

An issue for water

2011-04-08T12:53:00.001-04:00

Some other 3D indications

2011-04-08T10:23:00.001-04:00

The goal image is as follows.

Other messed results

A simple application for groups of curves to indicate 3D information

2011-04-07T10:39:00.000-04:00

Simple water ideas

2011-04-07T05:19:00.000-04:00

Basic ideas

2011-04-07T05:17:00.001-04:00

Starting curves and final curves

sparse

dense

(End directions make it ugly.)

Scales

2011-04-07T05:15:00.000-04:00

The direction and order map

Another testing for parallel creation

2011-03-29T08:10:00.000-04:00

More results from parallel creation

2011-03-29T08:07:00.000-04:00

Basic ideas IV: tessellations in parallel

2011-03-29T08:01:00.000-04:00

Failed

2011-03-28T12:15:00.000-04:00

For a very complicated curve, how to grow a valid curve?

TYPEIII: Testing

2011-03-28T12:08:00.001-04:00

Angle control:

Curve variations

A little global control

A little global control (90 degrees from the previous two testing)

I've done with testing the basic ideas.

2011-03-28T12:04:00.001-04:00

Tried:
1. randomly add a curve each time and build the tessellation.
2. after having a distribution, each place builds two curves and create a tessellation.
3. Create the first curve in a region and calculate the distance map. Then grow the next curve at the maximum distance value.

I am going to test some natural patterns:
fish scales, wings, flowers, leaves.

Then:
rivers, hair, weeds, barks.

Random tessellation

2011-03-21T12:44:00.001-04:00

I cannot control the partition spatially as I want now.

Testing for a Magnetic-curve tree

2011-02-28T09:38:00.004-05:00

The difference for the CAH stippling and your stippling

2011-02-28T09:36:00.003-05:00

the Mould's stippling

the CAH stippling

Apply leaf-like patterns to an image

2011-02-28T09:34:00.003-05:00

Randomly place leaf-like patterns

2011-02-28T09:33:00.003-05:00

Color each group to show the color contrast

2011-02-28T09:32:00.000-05:00

Group of spline curves to show the exclusion effects

2011-02-28T09:31:00.001-05:00

Not much improvement for mosaics

2011-02-14T12:48:00.004-05:00

Testing the automatic segmentation to control assignment. Should provide an interactive tool for color selection.

Stone textures

2011-02-08T15:33:00.003-05:00

Both, like stained glass

2011-02-08T13:57:00.002-05:00

Color each tile

2011-02-08T12:27:00.004-05:00

Thicken pixels with high gradients

2011-02-08T07:15:00.000-05:00

Thickness control

2011-02-08T07:06:00.002-05:00

Simple renderer

2011-02-07T15:04:00.004-05:00

Should work on after effects now

2011-02-07T10:21:00.004-05:00

It's sure it is good for hairy contents. I am not sure if those images will be easier cases for classical mosaic methods or not. I am thinking to use Barbara's examples to guide color choices and shape maps.

Different parameters for smoothness and sparseness

2011-02-01T14:25:00.003-05:00

Fixed most of doubly edges

2011-02-01T14:04:00.002-05:00

The results are without length and direction constraints as previous. Their object boundaries might be crossed sometime. Not sure which way is best to merge the length control and direction control.

Focus on the doubly edges

2011-01-31T15:26:00.003-05:00

Not easy to fix.Right now it is a hard connecting.

Colored effects

2011-01-30T01:45:00.001-05:00

Very sparse version

2011-01-30T01:24:00.000-05:00

Some improvement

2011-01-30T00:25:00.001-05:00

Building and tree examples

2011-01-26T01:46:00.002-05:00

Current process for previous results

2011-01-24T12:15:00.007-05:00

Better for images with strong direction. Compared with previous results, for images like Lena, it seems long slivers work better than small regions.Also it looks sometimes we should keep the curved lines.Should have density control.

-------------------------

with density

(CAH stippling)

(David stipples)

Goals for curves

2011-01-24T10:51:00.004-05:00

shuttershock.com

http://www.flickr.com/groups/gmic/discuss/72157624425214456/

Group of similar curves

2011-01-24T10:46:00.004-05:00

Patterns

2011-01-20T14:29:00.003-05:00

Magnetic curves

2011-01-17T12:30:00.005-05:00

Variations for S curves and shocks