## [Algorithms] error metric for computing lightmap required resolution

 [Algorithms] error metric for computing lightmap required resolution From: Rowan Wyborn - 2005-01-31 00:47:39 ```hullo, Im playing with a simple adaptive sampling algorithm for our light = mapper. With each generated light map, it trys downsizing it to various = resolutions, and then picks the final resolution based on an error = metric between the downsized lightmap and the original lightmap. Im = having a hard time tuning the metric to give reasonable results thou. = I've tried all sorts of variations of max pixel error, avg pixel error, = std dev of pixel error, but nothing seems to work well in a majority of = cases.=20 I was wondering, are there any established error metrics for comparing = images in this way? thanks, rowan ```

 [Algorithms] error metric for computing lightmap required resolution From: Rowan Wyborn - 2005-01-31 00:47:39 ```hullo, Im playing with a simple adaptive sampling algorithm for our light = mapper. With each generated light map, it trys downsizing it to various = resolutions, and then picks the final resolution based on an error = metric between the downsized lightmap and the original lightmap. Im = having a hard time tuning the metric to give reasonable results thou. = I've tried all sorts of variations of max pixel error, avg pixel error, = std dev of pixel error, but nothing seems to work well in a majority of = cases.=20 I was wondering, are there any established error metrics for comparing = images in this way? thanks, rowan ```
 RE: [Algorithms] error metric for computing lightmap required resolution From: Michael Smith - 2005-01-31 01:22:32 ```Hi Rowan, > I was wondering, are there any established error metrics for comparing > images in this way? In general comparing images is a very difficult thing to do automatically. The most common way is to use the signal to noise ratio, or in many cases the peak signal to noise ratio. However there are many cases in which a high PSNR will not produce an image that a human finds particularly bad. I would suggest for your case that maybe what you should be seeking to minimise is the max gradient between pixels. This should handle the case where you have a fairly sharp divide between light and shadow areas, which is where you most want a fairly high resolution. Michael > -----Original Message----- > From: gdalgorithms-list-admin@... [mailto:gdalgorithms- > list-admin@...] On Behalf Of Rowan Wyborn > Sent: Monday, 31 January 2005 11:50 AM > To: Gdalgorithms (E-mail) > Subject: [Algorithms] error metric for computing lightmap required > resolution > > hullo, > > Im playing with a simple adaptive sampling algorithm for our light mapper. > With each generated light map, it trys downsizing it to various > resolutions, and then picks the final resolution based on an error metric > between the downsized lightmap and the original lightmap. Im having a hard > time tuning the metric to give reasonable results thou. I've tried all > sorts of variations of max pixel error, avg pixel error, std dev of pixel > error, but nothing seems to work well in a majority of cases. > > > thanks, > rowan > > > ------------------------------------------------------- > This SF.Net email is sponsored by: IntelliVIEW -- Interactive Reporting > Tool for open source databases. Create drag-&-drop reports. Save time > by over 75%! Publish reports on the web. Export to DOC, XLS, RTF, etc. > Download a FREE copy at http://www.intelliview.com/go/osdn_nl > _______________________________________________ > GDAlgorithms-list mailing list > GDAlgorithms-list@... > https://lists.sourceforge.net/lists/listinfo/gdalgorithms-list > Archives: > http://sourceforge.net/mailarchive/forum.php?forum_ida88 ```
 Re: [Algorithms] error metric for computing lightmap required resolution From: Ivan-Assen Ivanov - 2005-02-02 15:02:03 ```One thing I would try is compare not the lightmaps after various forms of compression/resizing, but sets of equivalent screenshots of the final results of this lighting - e.g. a lit level, lightmap-only or fully textured. ```