Visualizing with Texture: Lessons from Puzzles

Over the holidays, we put out a couple jigsaw puzzles, to solve collaboratively or otherwise take a break from holiday mayhem. These are big puzzles, we did a 2000 piece cartoon puzzle and a 1000 piece photograph puzzle.

Completed 2000 piece puzzle.

With such big puzzles, different strategies are used — for example, a color-blind family member is more reliant on texture over color. Strategies for solving a jigsaw puzzle should be interesting to visualization researchers, because many of the tasks to solve a jigsaw puzzle are visualization-like activities:

Search: finding pieces of that match a subset of visual criteria, based on shape, color, text, texture, etc.
Locate: finding the place in the portion already solved to insert the new piece.
Identify: find a singular unique piece with unique criteria.

To do these tasks, we might use many different visual properties of the puzzle pieces:

A. Shape: The first step is to find the edge pieces and solve the perimeter. While shape is generally not considered preattentive by visualization researchers – it is for finding border pieces. Puzzle borders are straight, all other puzzle pieces are curvy or jaggy — meaning it’s visually preattentive to quickly find those straight edge pieces in a sea of curvy bits.

B. Similarity: Find pieces that share similar features. For example, in the cartoon, this included:
Text: Puzzle pieces with text on them;
Color: Pieces of a similar color (red), then given many red pieces, subdividing those into bright red (helicopter), soft red (stucco wall), red with brick texture (brick wall);
Stripes: green/chartreuse stripes (an awning); or black lines a regularly spaced intervals (pickets on handrails) — stripes are a type of texture;
Shape: tree branches (brownish branchy shapes) and leaves (a ragged zigzag on light green or dark green);
Blur: Interestingly, there is no blur in the cartoon puzzle, but int the photographic puzzle blur was a useful cue. The photo had a sharp focus at one depth with increasing blurriness at further/nearer depths. This was a useful cue for sorting pieces by depth in the scene.

Similar pieces, e.g. with text, by color and color variation, thin stripes, stripes repeating at regular intervals.

Categorization: Note that this similarity task is not trying to decode identity of a specific piece, rather it’s a categorization task, similiar to unsupervised clustering. In some respects, it differs from current data science approaches to clustering where the number of clusters needs to be defined upfront. Instead, the number of clusters is never defined, and is interactively refined throughout the puzzle-solving process.

C. Alignment: Given the similar pieces, use secondary cues to align pieces, e.g.:
– Slope/angle: the green/chartreuse stripes form an awning and the stripes slope at a gradually varying angle: this angle is used to locate adjacent pieces
Texture orientation: the bricks and mortar are in a regular pattern meaning pieces can be rotated to the correct orientation. Stone courses at regular intervals then help locate adjacent pieces.
Texture spacing: lines that represent a handrail are pickets. Pickets are spaced regularly. Two pieces may be adjacent if the spacing between the pickets within a piece match the spacing across the two pieces.

Aligning by texture and text: shadow cross-hatch, brick coursing, text, stripes with regular spacing or in perspective.

D. Content inspection: Near the end of the puzzle solve, the puzzle was largely solved except for highly detailed pieces without strong continuity of color/texture/shape between adjacent pieces (e.g. scenes with lots of little people). In this case content analysis was required and consideration of associations, e.g. a crowd of people shouting, a room full of many technical devices and so on.

E. Other strategies: Not all strategies are visual! One person’s strategy was a trial fit: if the color/texture is close, try to jam the piece in. If it doesn’t fit, no need to visually scrutinize the piece.

So What? Puzzle solving uses visual features such as shape, texture, texture orientation, texture pattern regularity, and blur (in addition to color). These tend to be used infrequently in data visualizations, but might have potential to be used more effectively.

About richardbrath

Richard is a long time visualization designer and researcher. Professionally, I am one of the partners of Uncharted Software Inc. I have recently completed a PhD in data visualization at LSBU. The opinions on this blog are related to my personal interests in data visualization, particularly around research interests related to my PhD work- this blog is about exploratory aspects of data visualization not proven principles.
This entry was posted in Data Visualization, Search, Shape Visualization, texture, Visual Attributes. Bookmark the permalink.

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s