Painterly Portrait Knowledge            Note v3                                  Steve DiPaola (ivizlab.sfu.ca)

These are work in progress notes of our "Painterly Rendered Portraits from Photographs using a Knowledge-Based Approach" research. System is written OO Java. We are currently on V.04 of the implementation (early results shown) but have not implemented all of the painterly model ( see sec 3) algorithmically yet.

Contact: Steve DiPaola < sdipaola @ sfu . ca >

Contents:
      1. Extended Abstract (SPIE paper)
      2. Early Results
      3. Notes on Painterly Model
      4. Painting Examples and Algorithmic Approach

1. Extended Abstract (SPIE paper): top

(in-progess for paper results examples)

Title: Painterly Rendered Portraits from Photographs using a Knowledge-Based Approach

Extended Abstract:

Portrait artists using oils, acrylics or pastels use a specific but open human vision methodology to create a painterly portrait of a live sitter. When they must use a photograph over a live person as the source, artists augment this process, since photographs have: different focusing - everything is in focus or the camera focuses in vertical planes; value clumping - the camera darkens the shadows and lightens the bright areas; as well as color and perspective distortion. In general, artistic methodology attempts the following: from the photograph, the painting must 'simplify, compose and leave out what’s irrelevant, emphasizing what’s important'. While seemingly a semantic goal, artists use known techniques such as relying on source tone over color to indirect into a semantic color temperature model, they also use brush and tonal "sharpness" to create a center of interest, lost and found edges to move the viewers gaze through the image towards the center of interest as well as other known techniques to filter and emphasize. Our work attempts to create a knowledge domain of the portrait painter process and incorporates this knowledge into a parameterized system that can create an array of painterly rendering output by analyzing the photographic-based input which informs the knowledge rules.

Our research system uses a multi-space parameterized approach to approximate a knowledge domain for painterly rendering of portraits. The knowledge domain uses fuzzy knowledge rules gained from interviews with oil portrait painters, data from the traditional portrait painter process combined with human vision techniques and semantic models of the face and upper torso. By knowledge domain we mean that system does not work at the pixel or direct sampled color level but instead acts on the semantic level a human painter function at: facial expression, facial planes, tonal masses, human vision, and color temperature control. The knowledge data is parameterized into an n-dimensional space of low level rules which can be accessed at different semantic levels. Non-professional photographic imagery of people's heads are used as input. Since photographs "value-clump" compared to human vision, our system approximates High Dynamic Range input by using three bracketed exposures of the same content (a feature on most commercial digital cameras) to insure a wide sampling of color and texture detail in the darks and lights of the photographs. The system uses a multi-layer stroke analyzer/renderer which perceives and lays strokes down with large masses first, progressively using smaller stokes and more detailed analysis. This approximates how painters squint at first to read large tonal masses and progressively add greater levels of detail over exposed paint from the layer before. Facial recognition techniques are used to populate the knowledge domain with semantic data of facial areas which mimics the painter process of orienting brush strokes in the direction of the surface planes (i.e the plane of a nose, or the circular muscle around an eye). The recognition system also determines center of interest which will affect brush detail and tonal detail in the final rendering. Final color choices of each brush stroke are based on tonal and color temperature rules combined with indirect input image color sampling. Color sampling of the source photograph(s) occurs as a pre-process along with other image analysis. This determines input attributes like dominate value and dominate color, both overall and locally for the body tone, half tone and shadow areas as well as a color temperature plan and a lost edges (i.e. tonal contrast) plan. These input attributes are used at stroke analysis time by the knowledge system to indirect into a unique but constructed palette for the specific final painting. The system also uses an initial configuration file that affects and informs the knowledge rules allowing for a range of painterly styles such as expressionistic, impressionistic and others. These initial configuration values effect the relationship between the image, the knowledge rules and low level attributes like brush size/length/orientation. The knowledge rules insure the painterly outcome emerges out of the semantic data of the source image rather than using simple templated pixel processing techniques like those found in painterly commercial image filters. The systems favors human vision over photograph input, by weighting the knowledge rules which are human vision based over the source imagery. This non-photorealistic rendering technique has applications in gaming, creative consumer digital photography, as well as video and film effects.

2. Early Results top

steve1.jpg
steve2.jpg

newer:
steve-27l.jpg
src1i3832.jpg
steve2-2c3-6-40.jpg
steve24a.jpg

contact researcher for newer results!

3. Notes on Painterly Model: top

These are the sorted/ordered notes taken from interviews and reference material (those notes are 5 times this condensed size) on the cognitive process of painting (in general - western) and sometimes specific to portrait (or compositions with a clear foreground/background element). We are in the process of committing these ideas to algorithms, some of the early results are using 1/8th of these ideas. It is an iterative process. (Warning this is a private site, images referenced here are for academic reference of an internal academic research group). (WARNING these notes go long).

Tonal

Tonal value (tone) refers to the light and dark in a painting or source image/sitter. Values are gray scale - no color. Colors in a painting come from the correct tonal value. Get the tone correct, and then with that tonal value get a color by the rules of color temperature, and other rules in the color section. Artists squint when they look at the sitter to see more in tones than colors.
Value is the relative likeness or darkness of an object.  Value creates dimensionality.  Every color has a value.  Every color can be lightened with the addition of a lighter color or white, or darkened with an additional color or black. These will give that color new values. Too many values cause the viewers eyes to jump around, failing to find the center of interest.  A five value scale is useful. Colors then will be described both their hue and one of these five values:  dark, middle dark, middle, middle-light and light. Or using portraiture terms):
Five (or 6) basic tone values See ToneValues

1)         Body tone (or light) – in line with the light source. - in direct light (usually warm colors)
2)         Halftone – where the light begins to turn – in between light and shadow
3)         Body shadow – turned away from the light source. – the darkest (usually cool colors)
4)         Cast shadows – dark tones that are caused by subjects standing in the path of the light source.
5)         Reflections – tones by light striking surrounding areas and bouncing back onto the subject.
6)         Highlights - Body tone that is directly hit by the light.

The first 3 are the most important.
There are nine degrees of values – all colors can be reduced to one of these nine values. See 9types-tones
So to paint a portrait first paint the head in 3 divisions of value – light, halftone and shadow and within these categories fit the 9 degrees of tonal gradation (light to dark). Not all paintings have all 9 degrees. It depends if it is a high-key, low key or middle key painting. Some painting systems use much less than 5 types and 9 values. A good technique however is to start by putting in your darks then halftones.
- Design is tops, Value comes next. Color is down below these in importance to the artists.

A painting must have a domain value, either it's light, or it's medium, or it's dark.
A small sketch is dominated by a light value, an Impressionist landscape is dominated by a medium value and a Rembrandt portrait is dominated by a dark value. The more you veer from this principle the weaker the painting. With 3 values: dark, mid and light, one must be dominant, the other two together will make up less than have of the first. -- none being of equal amount of the others. See Dominate&Subs

If you can't easily discern a paintings dominant value - dark, medium, light - then it fails that category. If a painting has great design than dominant value recedes to a second level of importance.

- Light areas (body tone) and dark (shadow areas):"neither one should borrow values from the other" As the face curves from the light, in general, the shadow side is 40 % darker than the sunlit side.

In a light painting beware of exaggerating the darks. When the dominant value is very light, we have a tendency to make the darks too dark. Best to moderate the dark in the shadows filling them with color that is seen there.

If you get the tone right in a stroke, you can almost pick any color (within that tone).

Color
The tonal value tells what color to pick within the other rules of portrait color. The most important of which is color temperature - that is warm colors and cool colors which work harmoniously with certain areas of the face ( the 5-6 tonal types but mainly body tone and shadow).
- Warm light appears more yellow, orange or red, while cooler light has a blue, green or violet in it. Light has its own color in it, so a warmer light we make green look yellow-green. Warmer colors come forward when placed near similar colors that are cooler or grayer. Foremost concern when picking a color is getting the value correct.

Hue is simply another work for color.  Intensity or saturation is the degree of strength or purity in a color.  Every color when mixed with its complement will give a rich complementary gray. Temperature refers to the relative warmth or coolness of any given color.  Colors can be mixed to create both cool and warm version.  The temperature of a color is always relative, i.e., green is warmer than blue (green contains yellow), but is cooler than yellow (green also contains blue).   Placing a warm color next to its cool version sets up a beautiful visual vibration.  It can support the center of interest or add energy. All portraits and painting exploit the relative positioning of warm and cool color stokes.

Determining the color of the light on your subject and whether it is in light or shadow at the first and most important color decision. Light give its temperature to what it touches. Everything left in shadow takes the opposite temperature. When the light areas are going to be warmed tones, generally speaking the shadows will be cool. Most portraits have warm light (skin) and cool shadows but there are many exceptions.

Warm lights produce cool shadows.
Cool lights produce warm shadows.     See WrmlightCoolShad
- The rule of unequal balance dictates that both light temperatures could not be shown equally. Usually within a portrait the warm in light is dominate over the cool in shadows. Also to make portraits work, temperatures along the center section of a face straddle the fence - cooler than warm light, yet warmer than cool light. See UneqWarmLight inversely subject lit by cool, indirect sun (or artificial) light have warm shadows.

The closer the subject to the light source the stronger the color will be. So a nose might pick up the light more than the rest. Again see UneqWarmLight

Once the color of light ( body tones ) has been determined , it is easy to come up with the color of the shadows.

Besides a dominant value most successful paintings have a dominant color. Having a dominant color simply means making it more warm or cool in a painting, in a figure painting you are more likely to making them warming hues.
- You can successfully save going overboard with cool hues with a couple of spots of warmth. And a hot canvas can be relieved by a small area of cool.

Placing equal amounts of warm and cool in a painting and equal amounts of light and dark can make a painting look weak and indecisive. Avoid your natural tendency to equalize.

Most system use the 3 primary color system, but a more scientific approach is the Munsell Color Notation system's 5 "principal" colors ( red, yellow, green, blue and purple) spaced equally around a color wheel.   First the standard 3 primary systems:

Make a decision about your portrait color harmony before you start. There are many color harmonies ( see bib Saper-p120) Here are two:
- Analogous Color Harmony for Portraits: includes adjacent wedges of color on the color wheel, including grayed-down neutrals and light and darker version of the colors themselves. Example is Yellow green analogous color harmony Skin orange ----- yellow -------green
- For portrait painting, orange will be used in addition to the analogous wedge.
- Example is Yellow green analogous color harmony:                         Skin orange ----- yellow -------green
- So pick the dominant color choice and then the wedge. Discords can be added in very small amounts at five o'clock and seven o'clock form the dominant (at  12). Use very sparingly discords give energy and capture the eye near you center of interest. See AnalogousColorHarmony
- Blue- green analogous color harmony is:   Skin orange ----------- | ----------------green ---------------blue
- Red-orange analogous color harmony is:    Red ----------------Skin orange 

P39 Complementary Color Harmony is the most versatile for portrait use. Since all skin has some aspects of red and yellow in it, you can easily adapt to a red-green, yellow-violet or blue orange scheme, if you keep in mind the principle of unequal balance. One of the colors in the portrait must be dominant one must be subordinate. Complementary ( is in the examples given here means the color directly across the color wheel ( red-green, orange-blue, yellow-violet   See ComplementaryColorHarmony

The Munsell Color Notation system's 5 "principal" colors ( red, yellow, green, blue and purple) spaced equally around a color wheel.   Munsell shifts the complements, so  red's complement to blue-green instead of green, yellow is purple=blue rather than blue.

A painting can not live by a dominant and compliment alone. We need to add more colors to create interest and excitement. The Munsell system offers a way of arriving at the best colors to use in this case.

 p. 71   For instance, the dominant color is the red family.  Now we add a complimentary color which adds contrast and interest but only a modest amount of it.  Blue-green is on the exact opposite side of the Munsell color wheel to red.  Just as white looks whiter against its opposite black so red is enhanced by its opposite blue-green.  The use of a complement would be destroyed by its overuse. This enhanced contrast effect works for any two complementary colors so yellow would look good with blue-purple. Discords are simply colors that are equidistant on the color wheel from the dominant hue and from each other. Locate them by an equilateral triangle on the Munsell color wheel.  Discords bridge and add visual excitement when used sparingly and in approximately equal amounts.  They are often located near the center of interest.

 See Munsell

Graying down a complement will allow you to use more of it.  Too much intense color can ruin a painting.  On the color wheel the closer a color moves towards the center the more the complement is introduced to it, graying and reducing its intensity. Summary 1) a painting should have a dominant color that is readily evident; 2) complementary color may be introduced, occupying much smaller space at full intensity, or greater area if grayed, 3) discord colors which add spice may be added sparingly in approximately equal amounts.

Shadow and reflected colors. The color of skin in shadows ( and lesser so in light) is influenced by the color of the background or clothing, in the turning down or away planes ( underside of chin and nose).

The laws of light dictate that horizontal surfaces reflect the sky and are therefore cooler (bluer) in color. Surfaces facing the sun will be warmer (yellowier) in color. Where sunny light and shadow meets is oblique to the sun therefore somewhat cooler. The Shadow side of the head is about 40 darker then the light side. So the flow of color and value is broken up by the facial features. The colors on the side of the nose and cheek receiving the sun’s rays directly tend toward orange. As the face curves from the light, the color cools and on the shadow side is 40 % darker than the sunlit side. It is only on the shadow side that we can see the glow of reflected light.

A contrast of cool and warm color is visually appealing. Often in portrait

Cool colors are used for:

Halftones Shadows Receding planes Highlights on black skin< Oriental skin tones< Veins in the hands

Teeth Whites of the eyes Solemn moods Older subjects Pale skin

Warmer colors are used for

Lights highlights  (except black skin) Lips Earlobes Knuckles Cheeks

Advancing planes An upbeat mood Younger subjects Ruddy or tanned skin

A color and its complement cancel each others identity and form a version of grey.  A color finds its identity and actually is intensified if it is put in juxtaposition to its complement.  A color’s intensity can be lessened, neutralized, or grayed by adding its complement.  Where light can strike color can appear in its full intensity.  Where light can’t strike the color diminishes such as the turning point of the body shadow and the edges of cast shadows.  The highlight is where the light strikes so hard that the color disappears in the glaring illumination.  This light is relatively complementary to the color it is on.  The body shadow is affected by reflected light so it has color but it is diminished in relation to the body tone.  This can be color toned down with gray or with its complement.  A reflection is a conditional color.  It depends on the conditions that cause it.
- Often rich warm color is exploited in "the core", like the shadows core of the length of the nose. Often this does not exist in a photo and must be put in. see Core
- As we look at distances in real life, yellow is the first color to be filtered out by the atmosphere. Reds go next. Blues last of all. That's why green hills and mountains are blue and purple in the horizon fading eventually to grays.

Shadows
- Form shadow is shadow on the side of an object.
- Cast shadow falls upon another surface as a result of something blocking the light.
+++ Cast shadows are generally darker and sharper-edged than form shadows, Form shadows typically has secondary light hitting them where cast shadows block light.
+++ Cast Shadows usually have a sharp edge.  A cast shadow usually becomes lighter and soft-edged the farther away from the object that is because the shadow is.

Shadow truths:
1)      There are no hard edges on the shadow side. All edges are muted.
2)      There are no textures/details in the shadows. They are flattened or blurred
3)      No strong colors in the shadows.
- The shadow edge - The more gradually an objects turns away from the light the softer and wider its shadow edge will be. A box has an abrupt turn and therefore its shadow is very sharp. So a forehead or cheek turns gradually- make a soft shadow edge -- a nose would be sharper.
- Hard edges are found in cast shadows, closest to the object casting the shadow.  Soft edges are fuzzy such as in form shadow.

Differentiating between the harder edge of a shadow cast by the nose on the upper lip and the softer edge of the form across the bridge of a nose or curve of the cheek will create energy.
- The shadow core - As a shadow goes from light to dark, the soft transition edge is called the "core". The core is where the color and value are the truest - neither obscured by shadow nor bleached by light. So the core is where the artist's hit there color most strongly. See ShadowCore
Avoid putting detail in both light and shadow areas if the attention is trained on the shadow side the light side should be more bleached out.

Once the color of light ( body tones ) has been determined , it is easy to come up with the color of the shadows. 3 shadow principles. 1) avoid strong color in the shadow ( gray down hues and lower intensity), 2) avoid hard edges in the shadows ( make them softer than those in the light) and 3) avoid strong contrast in shadows ( narrower range of values that in the light areas).
Shadows and reflects light. Areas where two objects ( say palm and cheek) touch in shadows have reflected light ( light bouncing back and forth). These makes an effect that causes a slight change in color and value, making the area of reflected light a tiny amount lighter that the rest of the shadow. It can be often overstated (because it looks so nice), it still should follow the 3 principles of shadow. Reflected light is darker than you think.
The edge that is created where the form turns and light meets the shadow is called the core of the shadow. Closest to the viewer the shadow core has 3 characteristics: the shadows core is darker in value that the rest of the form shadow, it is warmer in temp that wither the light or the shadow, and has the greater color intensity of either side. Again this might not show up in a photo but should in your painting. Again See ShadowCore

Composition and Overall
Shapes are more important than subject. Composition is more important than content. Link shapes to create pattern and thereby composition.
Creating movement with dynamic diagonals.  Lines and shapes moving vertically or horizontally convey formality and solidity. Diagonals convey movement and excitement.  Every subject has some diagonals.  A painter strengthens those elements and perhaps downplays the more static lines.
Overall portrait plan.
Determine your subjects:

            Center of interest.

                        Make an edge plan

            Your style of background.

            Your color harmony

                        Dominant color and color system (analogous / complementary)

To start the portrait I boldly lay down the light and dark pattern to represent the body tone and body shadow of the face, situating the large masses of the head.  Then I go in for more detail.  I start with the forehead and work down adding lighter tones to the body tone and dark to the shadow.

Reflection

Reflected light is not as light as it seems. Most artists overdo reflected light

Edge

Edges occur wherever shapes meet.  Edges accomplish 1) control of the viewers eye movement over the canvas – eyes always move to sharp edges and coast softly over soft edges.  On a lost edge the viewer finds comfort in seeking out the place where it is found again.  2) Sharp edges are at or in the center of interest.  Edges have four types:  hard, soft, lost and found.  Hard edges are found in cast shadows, closest to the object casting the shadow.  Soft edges are fuzzy such as in form shadow. Edges are lost when the shapes value is equal to the value of the shape next to it.  They are found again when one of the values change.  Due to the way a camera focuses, photos will give you false edge readings (sharp).  An edge plan can create a story from your painting.  For a viewer to find the center of interest, place sharp edges at or near this interest.  Sharp edges that have strong value shifts work even better.  Differentiating between the harder edge of a shadow cast by the nose on the upper lip and the softer edge of the form across the bridge of a nose or curve of the cheek will create energy.

Edges Lost and found edges

By softening or hardening edges or making them disappear entirely, the artist strengthens the illusion of form and gives a painting dramatic flow. In losing an edge we allow it to merge with an adjacent shadow, creating a link between objects - which is a powerful tool for design.  See Lost&Found
- The shadow edge - The more gradually an objects turns away from the light the softer and wider its shadow edge will be. A box has an abrupt turn and therefore its shadow is very sharp. So a forehead or cheek turns gradually- make a soft shadow edge -- a nose would be sharper.
- A subject can emerge from a strongly textured background, yet remain one with it, with the help of edges softened or entirely lost, shadow cores, lines that play hide and seek with your eye.
The same kind of brushstroke are used in the background as in the portrait but the BG the edges are softer than on the model

For the lost edge, pull the brush away from the canvas.  Start your stroke with pressure where you want a found line.  You need found “sharp” and lost “fuzzy” lines to make the planes of the entire periphery project and recede.

Center of Interest

As a painter you must choose your center of interest -- where you establish edge quality of your portrait. Make edge decisions based on your center of interest you pick. Sharpest edges at the center of interest, less sharp edges to move the viewers eyes across the canvas to the center of interest and softest edges where you want your viewers eye to glide. See CenterOfInterest

For a viewer to find the center of interest, place sharp edges at or near this interest.  Sharp edges that have strong value shifts work even better.

Create the center of interest. In your painting you’ll want to soften edges and reduce contrast, brightness and detail outside of what will become the center of interest. 

Leave out intense color that will abruptly lead the eye away from the center of interest.

Photography

From your photograph to your painting you must simplify, compose and leave out what’s irrelevant, emphasizing what’s important.  Create the center of interest.  A camera has none.  A camera will focus on everything.  In your painting you’ll want to soften edges and reduce contrast, brightness and detail outside of what will become the center of interest.  Photographs will make shadows too dark.  So they must be filled with subtle color and detail. Remember however that you don’t want to put detail in both shadow and light areas.  The same is true for overexposed areas. Due to the way a camera focuses, photos will give you false edge readings (sharp).

The 3 main problems with portrait painting from photography are: depth of field (everything is in focus or the camera focus in unnatural vertical planes - no center of interest focus) value clumping ( the camera darkens the shadows and lightens the bright areas) and color distortion (as well as perspective distortions).

focus - Our eye uses a 3D spherical focus, with things getting focused as you move out form the core. As a painter you must choose your center of interest -- where you establish edge quality of your portrait. Make edge decisions based on your center of interest you pick. Sharpest edges at the center of interest, less sharp edges to move the viewers eyes across the canvas to the center of interest and softest edges where you want your viewers eye to glide.

Value clumping - photos distort the tonal range, so values on either end of the scale are compressed and show less differential than actually exists in nature. Extend the middle value by extending other areas that look dark in value, lighter and look light, darker in value. See FalseDarks

Highlights

Highlights need color too. The very last things painted on a surface are the highlights reflecting the light source itself. The most effective device to convey the intensity of this is to change the temperature of the light source, ie warm light sources yielding cool highlights and visa versa. Highlight can be overused - they should have a location shape and color and should not be painted if they are not there. Highlights are the one things that will move with the viewer (different painters painting the same model will have different highlights), so they give away camera's eye.

Strokes

Never completely finish edges and be careful of too many highlights.  Always overlap rather than putting objects in a row.

When making brushstrokes think of the way the surface is going and paint in color to suggest the surface itself. Do not always brush the colors in the same direction, but let the form guide your brushstrokes. The right kind of brush stroke can create depth and dimension. See Value&Direction

Stroke your paint on in the same direction as the plane.  If you want the lower lip to protrude, pull your stroke out.  If you want the front plane of the cheek come forward and the side plane of the cheek to go down, actually apply the paint in this fashion.  There are shapes that follow vertical planes and those that follow horizontal.  But the planes from front to back give the dimensional effect.  Pull or push the stroke in the direction of the plane.  Strokes put down in the same direction become monotonous.  They have to be broken with strokes in opposing directions.  One stroke in one direction butted up against the next stroke in a somewhat opposite direction makes a good painting.  So balance and variety seem to be the key.  Stroke to match the shape.  Then affect it with a stroke opposite to it. See StrokePlanes

Choppy strokes – using a choppy aggressive technique for the clothing and background contrasts with the painterly look of the face. 

For the lost edge, pull the brush away from the canvas.  Start your stroke with pressure where you want a found line.  You need found “sharp” and lost “fuzzy” lines to make the planes of the entire periphery project and recede.

Background

The background affects the overall relationships and should be painted first or at the same time. The background is unique to portrait artists. The main rule of background is that they recede. 4 main backgrounds:

1 the plain solid colored background
2 the imaginary background – made-up of shapes and colors
3 a representational background of what is really there
4 an abstract of what is really there.

For 1 you can use a color in your portrait to maintain color harmony. The inverted background is one of the most common

The same kind of brushstroke are used in the BG as in the portrait but the BG the edges are softer than on the model

Busy backgrounds confuse the image unless they use an unrelated color to the subject.  Warm backgrounds bring out warm flesh tones. A cool background emphasizes cooler tones.  A light background shows up the shape of a dark figure.  A darker blends with dark tones on the subject.

Mood

Dark pictures lend themselves to a quiet, more contemplative mood.

Warm light lends a restful, comfortable mood to your portrait.

Images requests:

High key, mid key and dark portraits

Warm light and cool light images

Examples:

ex_ramsey
ex_sutherland
ex_drut
ex_rodwellCU
ex_hortonCU
ex_hortonReal
ex_sandonCU
steve1.jpg
steve2.jpg

4. Painting Examples and Algorithmic Approach top

(this section is outdated !!!) see authors for current code and research direcion
(Warning this is a private site, images referenced here are for academic reference of an internal academic research group).

col
gross (3)
fine (18)
ojb+coi

ff-col
ff-painting
ff-fine (9)
ff-ojb+coi

painter brushes

Algorithm

- Pre distortion / pre crop
- Create 3 tone image GTmap ( GrossToneMap) [
- Create object matte - OBJmap (describes objects: background, hair, cloths, non-face skin,

Based on Source Image define parameters: (can by done by human or via automatic analysis)
       - Center of interest- COImap
              - hand picked could be a point, full matte or matted area
       - Dominant Value & Value Plan
               - via analysis of source, but human choice
       - Dominate color & Color Temp Plan & Palette (initially just Palette)                - via analysis of source, but human choice
       - Edge plan
               - unknown how to pick or use yet
       - Brush Style
               brush type
               brush path type
               full brush direction image?

    System take inits data and params:
        Create 9 tone map (FTmap- Fine Tone Map)
        Upgrade tonal map to take into account shadow core/ clumping and other photography issues
        Use GTmap to iterate brush size levels ( main loop)
        Use OBJmap + COImap + TONmap to dictate 'sharpness' at different levels
        Use  above + ValuePlan + ColorPlan  to dictate palette choice at different levels
    Use above + Brush Style  to dictate brush path/size/type  at different levels      
    Rather than just gradient brush direction, a more sophisticated brush render is needed.
    Some sense of neighborhood is need for values and brush

Goals:
    Creating a system not an algorithm, meaning the parameterized system has depth in the kind of paintings styles it can make via changing the parameters and the source image.
    Resolution independence - the system will work at the resolution given, fast at low res and detailed at high res.

additional info:

HertzmanNotes