From: Mitch S. <mit...@be...> - 2010-03-20 01:55:06
|
CHAN, KENNETH 1 [AG/7721] wrote: > Hi guys, > I encountered a strange result when using wig-to-json.pl. The > quantitative track I got has a very low resolution and the values > doesn't seem like corresponding to the wiggle file. Hi, I haven't gotten a chance to reproduce this or look at it in detail yet, but I do have a quick answer to part of this: Currently, JBrowse renders wiggle files up to one pixel per base (because that's the highest possible data density) and then uses the web browser to stretch the images when the user is zoomed in closer than one pixel per base. On large genomes, this saves a lot of disk space and processing time compared to rendering images at higher resolutions. Plus, JBrowse allows for variable font sizes, so at the highest zoom level, different users will have different values for the number of pixels per base. So even if we wanted to generate higher resolution images, we wouldn't be able to pick a scale factor for the highest zoom level that would work exactly for everyone; in many cases the web browser would still have to stretch or shrink the image. Originally, the web browsers all used a scaling algorithm that kept the results crisp-looking, but more recently they've started to use a smoother (bicubic) scaling algorithm by default, and wiggle tracks started looking fuzzy in JBrowse when they're zoomed in all the way. The web browsers have just started implementing switches that will allow us to specify the old scaling method: https://developer.mozilla.org/en/CSS/image-rendering and writing the JBrowse code to set that value appropriately is on our to-do list (although if anyone wants to tackle it, that would be welcome; it should be simple to set the relevant CSS styles on the images in ImageTrack.js). Mitch |