tag:blogger.com,1999:blog-337536679221777590.post7853497786880856387..comments2011-06-16T04:18:18.090+01:00jfshttp://www.blogger.com/profile/13154641479666334229noreply@blogger.comBlogger5125tag:blogger.com,1999:blog-337536679221777590.post-62202466825058242002011-06-16T04:18:18.090+01:002011-06-16T04:18:18.090+01:00Very useful. Thanks!Antoniohttp://www.blogger.com/profile/17709582966428668709noreply@blogger.comtag:blogger.com,1999:blog-337536679221777590.post-47957529950680957922009-06-22T10:37:19.242+01:002009-06-22T10:37:19.242+01:00Please ask your support question in the forums, not in a blog comment.jfshttp://www.blogger.com/profile/13154641479666334229noreply@blogger.comtag:blogger.com,1999:blog-337536679221777590.post-55535161035257815812009-06-21T23:08:54.738+01:002009-06-21T23:08:54.738+01:00I put {\be10} in my video, and i when i encode it with x264 it doesnt show :( What is the problem? I also use K-Lite codec pack. Can someone please help me out?Anbuhttp://www.blogger.com/profile/11007284136873587321noreply@blogger.comtag:blogger.com,1999:blog-337536679221777590.post-68320706057659106162008-10-12T22:52:00.000+01:002008-10-12T22:52:00.000+01:00Heh, late reply (somehow hadn't noticed this comment before) but here's an answer anyway.<BR/><BR/>The \be tag uses a simple 3x3 matrix, [ 1 2 1; 2 4 2; 1 2 1 ]. It is not applied in a separated manner but just as a 3x3 convolution.<BR/><BR/>With the "extended \be" patch you can give a larger parameter than 1 to \be, this causes the convolution to be applied that number of times. As the image in the main post shows, the results aren't very pretty.<BR/><BR/>The \blur tag uses a real gaussian kernel applied as a separated filter. It is an approximated kernel, the elements are unsigned bytes ie. 0-255 range, and rather than being of infinite length it uses a cutoff which should be large enough to get the significant part of it.jfshttp://www.blogger.com/profile/13154641479666334229noreply@blogger.comtag:blogger.com,1999:blog-337536679221777590.post-18622742402882990192008-08-21T00:52:00.000+01:002008-08-21T00:52:00.000+01:00Another enjoyable article! <BR/><BR/>I guess the blurring employed by VSFilter runs a weighted kernel across the bitmap. Does it use a source and clone bitmap for this or just the source? The results can be quite different. What size kernel is used? 3x3? 5x5? Or does this depend on the user-selected depth? In other words, if it uses a 3x3 kernel (blur = 1) and the user wants a blur depth of 2, does it run the 3x3 code twice or increase the size of the kernel to 5x5? Is the blur process split into two passes (vertical and horizontal) to increase rendering speed? And is true Gaussian blurring used or just an approximation?<BR/><BR/>Sorry for all the questions! I find this stuff fascinating, having knocked myself out for 24 months adding a pseudo-Gaussian blur to our app's bitmap subtitles. And I'm still not happy with the results! (http://www.spotsoftware.nl/bitmap.html) :)Mark Raishbrookwww.spotsoftware.nlnoreply@blogger.com