You are here: Home Chapter 6 - Exploiting Weave Maps 6.5 Concluding remarks
Document Actions

6.5 Concluding remarks

Thirty percent (10 of 34) of Vermeer’s paintings are on canvas that has a weave match with another painting in that group of 10. This represents a significant fraction of his extant oeuvre and as such is a common part of his canvas acquisition and preparation procedures. Weft snake indicators appear in fifteen of his 34 paintings, more than 40%. How many actually locate weft snakes? Several do. Over half of his paintings (18 of 34) on canvas have cusping over 5 cm in depth on all four sides. What is the history of the ones that do not? We have considered some possibilities suggested by our canvas weave examinations.

Weave maps have led us to new knowledge, and in the process raised the prospect of being able to answer even more penetrating questions. Visualization tools being developed for the study of artworks, such as for the Bosch Project, will expand the utility of weave maps.1

The next generation of automated thread counters is arriving, including sophisticated improvements on spectral-based methods2 and procedures locating every thread intersection.3 These approaches are enlarging the range of image quality and thread thickness variability for which automated thread counts are quite close to local hand counts. Still, aggregation will be needed in creating weave maps that exhibit the stripes used for weave matching. It is an issue of scale and texture. Consider the size of the patch of the image over which the local thread count is evaluated. The bigger the patch, the more threads in the computed average. The more threads in the average, the more their average matches the true average of the whole set of patches. With too few threads in the average relative to the range of thicknesses encountered, the average counts in the patches range too erratically. In neither extreme do the stripes appear. This can be affirmed by experiments using displayWeaveMaps with different evaluation square sizes. The proper scale depends on the particular piece of canvas and the image quality. See Figure 5.9a and Figure 5.9b in the previous chapter for an example.4

The new insights gained from the development of automated thread counting software and the visualization of its output as weave maps make a compelling case for more oeuvre-covering, scientific quality image datasets of X-radiographs of paintings by other artists and groups of artists. A natural extension of the Vermeer dataset underpinning this RKD Studies publication would be to expand it to cover seventeenth-century Dutch painters working in Delft. This groundbreaking example of computational art history quickly scales up to a ‘big data’ issue with its attendant complexity and promise.



[1]

 http://boschproject.org/ (date consulted: September 27, 2017)

[2]

 H. Yang, J. Lu, W.P. Brown, I. Daubechies, and L. Ying, ‘Quantitative Canvas Weave Analysis Using 2-D Synchrosqueezed Transforms. Application of time-frequency analysis to art investigation’, IEEE Signal Processing Magazine 32 (July 2015), issue 4, pp. 55-63.

[3]

 L. van der Maaten, and R. Erdmann,Automatic Thread-Level Canvas Analysis’, IEEE Signal Processing Magazine 32 (July 2015), issue 4, pp. 38-45.

[4]

William A. Sethares, 'Automated Creation of Weave Maps’, in: C.R. Johnson Jr. (ed.), Counting Vermeer: Using Weave Maps to Study Vermeer's Canvases, RKD Studies, The Hague (RKD) 2017, (countingvermeer.rkdmonographs.nl/chapter-5-automated-creation-of-weave-maps).

Datum laatste wijziging: Oct 16, 2017 01:38 PM