Browse Prior Art Database

Automatic legend layout and location design in figures Disclosure Number: IPCOM000243015D
Publication Date: 2015-Sep-09
Document File: 2 page(s) / 68K

Publishing Venue

The Prior Art Database


This diclosure proposes an automatic method for designing the position and layout of the legend, which is associated with the curves or plots in the figure. A fitness function is designed and evaluated over the candidate position in the figure in a sliding-window fashion. This search process returen the optimized position and layout of the legend. The fitness function involves acriteria reducing the uncertainty due to the occlusion by the legend, as well as the overall unity of the legend layout across multiple figures.

This text was extracted from a PDF file.
This is the abbreviated version, containing approximately 68% of the total text.

Page 01 of 2

Automatic legend layout and location design in figures

Curve figures are ubiquitous in business intelligence and other analytics scenario. Legends are often accompanied with curves, and important to readability and attract readers' interest. In many publishing report/papers, it is too space-wise luxurious to attach the legends outside the figures and may bring about additional ambiguity. Or sometimes just want to make the layout more compact, and it is costive for readers to cross-move their eyes here and there.

The following figure gives an example of ideal and optimized legend layout associated with the dense curves.

Our approach consists of the following steps:

1)Create a legend template by the font, format and content


Page 02 of 2

The font and format can be determined by the user, format refers to the orientation of the legend etc.

The content corresponds to the curves in the figure

2)Traverse the whole figure by a fixed step size in a sliding window fashion

3)In each candidate location, compute the fitness score by

For the area hidden by the legend, perform interpolating techniques to recover the curves

The interpolating models can be

User pre-configured, such as the linear model, quadratic model, etc.

Data-driven model, which use the complete curves in the dataset to recover the hidden part, e.g. using similarity search to find most similar curves or segments, and interpolate the missing parts with other samples

Compute the error between the recovered cu...