User:Thijshijsijsjss/Pen Plotting Panache/Plothatching Workflow

A documentation of my workflow for making multicolor crosshatching plots. It should (ultimately) be comprehensive for anyone by links and step-by-step guides, but is for now primarily written for my own future reference. This means that there will be many options for each of these steps, and I will limit myself to those strategies I have found to work for me.

I will make reference to several pieces of software. Notably:

• Krita, open source painting software (V5.1.5)
• Inkscape, open source vector manipulation software (V1.2)
• Some code I have written for crosshatching. Find it here on GitHub. Also, find its dedicated wiki page here

Image selection

There's no real limit to what images can be plotted! These techniques are pretty flexible and powerful. However, still image selection seems to be one of the harder steps in this process.

Some things to keep in mind:

• Plotting takes quite a bit of time and attention. Not only the act of plotting itself, but also the preparations require a significant amount of time. Choose your current project accordingly.
  • Monochrome plots are way faster than multicolor plots, both due to reduced plotting and prepping time. Additionally, the babysitting that tends to be automatic when penplotting is largely only important when switching out pens. One pen (one color) can be safely let free to plot without being babysat.
• Your selection of pens matters. Just like every plotter has its quirks, every combination of pens has its unique vibe. It's not a bad idea to take your options of pens into consideration for your choice of image.
• The crosshatching technique tends to work best for certain images:
  • Images with high contrast; this will improve the level of detail that will be perceived in the final plot
  • Multicolor images where the different pens will need to interact 'interestingly': nice pieces of overlap balanced with areas of one color.
  • Multicolor images don't want to be very dark when plotted with a dark color for 'key'; the 'key' color will just cover everything

Image preparation

Dimensions

Example of 500px (bottom) and 600px (top) plot widths

Example of a plot of 1000px plot in width

To avoid having to scale many layers individually later on, it's easiest to determine the size of your plot at the start. As the plotter is fastest in its horizontal axis, it is advised to orient your image in a way so that the largest dimension is the horizontal one (i.e. by rotating a portrait mode image). However, when wanting to make a plot as large as possible, this is obviously not always desired. In that case, keep bounds of 1100px by 1450px in mind (the full size of the paper in pixels according to Inkscape's default conversion).

My three most used choices for image width are:

• 500px: This allows for two same width plots to be placed next to eachother on the width of the A3 plotter I have been using.
• 600px: 500px plots feel quite small. I have found 600px to be the sweet spot for compact plots that still capture quite some detail.
• 1000px: this is the about the full width of the A3 plotter I have been using.

Note that this conversion between pixels and real world measurements units is actually arbitrary. This discussion has the Inkscape default conversion in mind. Relying on this conversion saves us potentially difficult conversion problems later on, and has proven reliable between plots.

When planning to plot an image not its original orientation, this is the ideal time to apply the appropriate rotation

Passe Partout

Eventually, we'll need to create vector paths out of the images the crosshatching software outputs. As plotting is a lenghty step, any preparations we can do to improve the plotting time, will reduce the overall time it takes to create a plothatched image. One such improvement is the way the vector paths are traced. In Inkscape, 'Centerline (autotrace)' has the best results, by far. However, it will often (~60% of the time) crash Inkscape, in which case one has to resort to less efficient tracing methods that will typicially cause each line to be drawn from it's center (effectively adding >50% extra plotting time).

By adding a white border, a 'passe partout', around your image, Centerline (autotrace) has a better chance to succeed. In fact, borders of 100px have had a 100% success rate for me (n>150). This will NOT affect the final dimensions of your plot.

1. Open your file in Krita
2. Go to Image > Resize Canvas
3. Add 100px to the smallest of your image's dimenstions
4. Click OK
5. Save image (overwrite it)

Layer separation

We will separate the different color channels and save them as individual files, so that we will be able to plothatch them individually.

1. Open your file in Krita
2. Change color space to CMYK
   1. Image > Convert Image Color Space
   2. Choose Model: CMYK / Alpha
   3. Click OK
3. Separate image
   1. Image > Separate Image
   2. Click OK
4. Invert all layers. For each of the layers:
   1. Go to a layer and press ctr+I
5. Save layers 'cyan', 'yellow', 'magenta' and 'black' separately. For each of the layers:
   1. Toggle visibility for only one of these layers
   2. Save (ctr+shft+S)
6. Fix opacity. For each of the new images:
   1. Open the image with the Windows default images app
   2. Three dots > Change image size
   3. Save image (overwrite it)

Transparancy adjustments

The previous step will typically output jpg images for which the transparancy is interpreted as black. Often this is not what we want. Magically, opening an image in Windows's default image viewer and re-saving it (overwriting it) fixes the transparancy. This is a quick process. For other operating systems, I imagine Krita and Inkscape both support functionality to handle this.

Hatching

This step describes the conversion of your selected and prepared image file to crosshatched layers using some Processing code. A dedicated page of this software can be found here.

First time only: downloading the script

To crosshatch your own images, you'll need to be able to compile a Processing file. For this, there are several options:

• The easiest is downloading the Processing IDE
• You can also compile from the command line
• Your favorite IDE might offer plugins for Processing language and compilation support, like this one for vscode

Next, you need to obtain the code. Again, there are several options:

• Go to https://github.com/Nyxaeroz/Crosshatching, press the button that says code, download the zip and unzip it wherever you want this project to reside on your machine.
• Moving to your desired location and use the following command in a terminal to clone the repository: git clone https://github.com/Nyxaeroz/Crosshatching.git

Preparing script

The script uses some parameters. Crucially, you need to specify the dimensions of your image in two places.

1. Open sketch.pde
2. Change the variable input_name (line 4) to match your input file's name (if this file is not located in the same folder as hachter.pde, provide the path).
3. Change the variables input_width (line 11) and input_height (line 12) to the dimensions of your file.
4. Change size(W,H) (line 18) such that W and H are the width resp. height of your file.

Optional Parameters

Several parameters may be altered to achieve different results. In particular:

• int nr_layers (line 6): the number of crosshatch-layers to be used.
• int v_line_offset (line 8): the space between two lines within one crosshatch-layer
• float blur-radius (line 9): because, often, we don't want the smallest details to appear in our crosshatch, we apply a Gaussian blur. This radius determines it's size. A size of 0 can be used to not use a blur.
• boolean save (line 14): the program will save the layers if and only if 'save' is set to true. Not saving would still display the image, and is useful e.g. in tweaking the parameters.

I have found that a default configuration of nr_layers=4, v_line_offset=5 and blur-radius=2 has a good success rate. I usually set save=false to try out some tweaks before setting it to true to save.

Hatching each layer

If you've followed the steps above, all images for the different color channels should be of the same dimension. This is useful, as this means preparing the script needs to be done but once (aside from choosing a different input_name of course).

For each layer, the optional parameters may be tweaked individually. By default I use the same configuration for every layer, however not uncommon exceptions include:

• using fewer layers for the Key color channel, as this layer tends to dominate the final result.
• using just one layer for big fills of on color (especially if I know I'm using markers or other wide tipped pens).

For each color channel you want to hatch:

1. Set input_name (line 4) to the corrent filename.
2. Tweak the optional parameters if so desired
3. Run the program
4. Find the output files, seperate svg layers, int a folder output_layers
5. Move these layers to a new location so that they are not overwritten.

Customizations

Manual thresholds

Sometimes it is useful to manually set the threshold values for each layer. For example when aiming for specific effects, or when an image's range of values is small. Add these lines on top:

boolean use_manual_t = true;
float[] manual_t = {0.1,0.2,0.3}; // specify nr_layers many numbers

And change line 36 to:

if ( use_manual_t ) {
  output_layers[i] = create_layer( manual_t[i], my_theta );
} else {
  output_layers[i] = create_layer( my_thresh, my_theta );
}

(Similarly, one could use manual angles for each layer. I've not explored this yet, but there's some interesting potential here, e.g. in incorporating Moire Patters or playing with optical illusions).

Custom output folder

Setting custom output folder: just like the input_name (line 4) is explicitly specified, one can explicitly specify an output location. Add this line on top:

String output_folder = "MY_OUTPUT_FOLDER";

And modify line 41 to be:

beginRecord(SVG, output_folder + "/layer_" + i + ".svg");

Inkscape

Prepare document

We will prepare the document we will be working with. First, set the documents size:

1. Go to File > Document Properties (Ctrl+Shift+D)
2. Choose Format: A3 (or any other appropriate filesize). Note that this is primarily for the width, and not so much for the height of the paper (i.e. feeding a sheet of A4 paper in landscape mode to the plotter and plotting something from A3 settings that fits within the bounds of the A4 sheet will not cause any problems).

Next, create the layers for each color channel, and one additional layer that we'll use for importation and organization purposes. In the case of a CYMK plot:

1. Open de Layers and Object tab by going to Layers > Layers and Objects (Ctrl+shift+L)
2. Create 4 new layers with the dedicated button, or with Strl+Shift+N
3. Rename the layers res, cyan, yellow, magenta, black (this is not strictly necessary, but useful for referencing layers within this guide, and also useful for yourself later on).

Steps for each color channel image

We will now do some preparations for each color we'll be plotting with. Confusingly, there are two notions of 'layer' here: the Inkscape 'layers', which can be thought of as folders that contain all the paths that sould be drawn with one pen. Then, there are the 'layers' of the physical world: layers of the same color that will be drawn on top of each other.

The following steps need to be taken for each color we're going to plot with, i.e. for each set of output layers we have created with the crosshatching software.

1. Select the res layer
2. Import all crosshatched layers of one color channel to the res folder. This is easiest by selecting them all in a file explorer and dragging them ONTO THE CANVAS AREA (note that Inkscape will crash when they are dragged onto the layers menu), but this can also be done one by one through Inkscape's import menu.
3. For each imported object, you'll see a sublayer in Inkscape with an image. For each of these sublayers:
   1. Position it at (0,0) by selecting the Select Tool, selecting the relevant sublayer, and setting x and y fields to 0.
   2. Still with the Select Tool, select the image in the sublayer, and go to Trace Bitmap (Shift+Alt+B, Path>Trace Bitmap, or through the quick menu if it's still open)
   3. MAKE SURE SMOOTH CORNERS IS TICKED OFF
   4. Click Apply
4. Move all paths that have been created to the appropriate Inkscape layer.
5. Delete all contents from the res layer.

Alignment thingies

When plotting through Inkscape, the top-left corner of your plot is determined at the time of starting the plot, and all paths in that 'batch' are positioned relative to that origin point. This means that when plotting two paths with top-left corners at (X,Y) and (X+n,Y+m) (n,m>=0), these will be plotted at (0,0) and (n,m). In other words: the relative position is maintained when plotting multiple paths at a time.

When sending new paths to the plotter, however, this origin is redetermined. So when plotting a path with top-left corner at (X,Y) and then separately plotting a path with top-left corner at (X+n,Y+m) (n,m>=0), these will both be plotted at (0,0). In other words: the relative position is lost when plotting paths separately.

Typically, we desire the relative position to be maintained. As it is always the top-left corner that determines the position, we will add a marker -- and 'alignment thingy' -- in the same absolute position for each batch of paths we will plot.

1. In Inkscape, determine the absolute position of your origin point. Let's call this (N.M)
   1. If you're not using a passe partout, (N,M) = (0,0)
   2. Otherwise, it will be different. For a passe partout of 100px both to width and height, (N,M) = (26,26)
2. Select the Square Tool and draw a square of any size on any position on the canvas (typically I would do this in the res layer)
3. Select the Select Tool and select the newly drwan square
4. Set its parameters: x=N, y=M, w=0.1, h=0.1
5. Use Right Click > Duplicate on this alignment thingy, such that you have one for each batch (typically: each layer) you'll be plotting
6. Move one alingment thingy to each batch (typically: each layer).

Tip: the visible layers will be plotted top to bottom, and the paths in within each layer bottom to top. Therefore, it is advised to place the alignment thingy as the bottom path of each layer. By doing so, it immediately becomes clear if there are alignment issues.

Usually, I'd go over each layers and set their position to (0,0). This is not relevant for the actual location of the plot, but is useful when wanting to position multiple plot on the same sheet of paper. On that note:

When plotting multiple plots next to each other on one sheet of paper, the alignment thingy can be used for the appropriate offsets. E.g. when plotting two 500x500px plots side by side, the latter's alignment thingy should be offset by (-500,-500).

Optionally, scaling and rotation

Sometimes you might want to scale or scale a prepared image, e.g. in an effort to use most of your sheet of paper, or to fit multiple plots side by side.

Ideally, rotations and scaling are done in the setting dimensions step. 

This is not always possible, however, e.g. if you're using the preparations of an image you've previously plotted in a different orientation. In this case, this is the time to apply these transformations.

Scaling is easy. For each layer:

1. Go to Object > Transform (Ctrl+shift+M)
2. Go to the scale tab
3. Input the new scale

Rotation is more involved. On the surface, it is similar. For each layer:

1. Go to Object > Transform (Ctr+shift+M)
2. Go to the rotation tab
3. Input the new rotation

HOWEVER: note that the origin of rotation is relative to the object. This means that properly aligned objects of different sizes might lose their relative positions after each rotation the same amount. Alignment thingies can be used to prevent this. Depending on your rotation, you might need to add an alignment thingy to another corner. Worst case scenario, you need to move your rotated object (or use Ctr+Z to undo).

Optionally, automated layer names

There is neat functionality within Inkscape to tell the plotter what pen to use and with what speed to plot. This can be done by giving a layer a specific name:

Pen {1,2,3,4,5,6} Speed 0

Exactly one number out of {1,2,3,4,5,6} should be chosen. For the speed, any natural number (or positive real?) can be used, but I've yet to come across a situation where I did not want to use 0 (fassst plotting).

Setting these layer names is very useful when wanting to automate a bigger plot: by naming all layers using this convention, you can plot with (in the case of my plotter) 6 pens 'at once'. This is the most reliable way of good relative alignment between layers.