Comments for Computer Vision For Dummies http://inperc.com/blog2 Computer vision, image analysis, and related mathematics Thu, 17 Feb 2011 13:25:51 +0000 hourly 1 http://wordpress.org/?v=3.1.3 Comment on Google confuses topology and topography by Peter http://inperc.com/blog2/2011/02/14/google-confuses-topology-and-topography/comment-page-1/#comment-878 Peter Thu, 17 Feb 2011 13:25:51 +0000 http://inperc.com/blog2/?p=643#comment-878 Yes, and it make more sense in this context - as one of several meanings of the word. It shouldn't be #1 though! That's what messed up Google. Yes, and it make more sense in this context – as one of several meanings of the word. It shouldn’t be #1 though! That’s what messed up Google.

]]> Comment on Google confuses topology and topography by Hesham http://inperc.com/blog2/2011/02/14/google-confuses-topology-and-topography/comment-page-1/#comment-877 Hesham Thu, 17 Feb 2011 13:14:46 +0000 http://inperc.com/blog2/?p=643#comment-877 It's from here: http://wordnetweb.princeton.edu/perl/webwn?s=topology It’s from here:
http://wordnetweb.princeton.edu/perl/webwn?s=topology

]]> Comment on Stereo vision with Kinect by Peter http://inperc.com/blog2/2010/11/15/stereo-vision-with-kinect/comment-page-1/#comment-872 Peter Wed, 17 Nov 2010 16:50:59 +0000 http://inperc.com/blog2/?p=570#comment-872 I stand corrected. Kinect has is a depth sensor. So, the link isn't relevant here. Thanks! This is still "stereo vision", I think. If this is a step towards making 3D imaging as common as 2D, that would be exciting... I stand corrected. Kinect has is a depth sensor. So, the link isn’t relevant here. Thanks! This is still “stereo vision”, I think.

If this is a step towards making 3D imaging as common as 2D, that would be exciting…

]]> Comment on Stereo vision with Kinect by edward_dixon http://inperc.com/blog2/2010/11/15/stereo-vision-with-kinect/comment-page-1/#comment-871 edward_dixon Wed, 17 Nov 2010 15:22:25 +0000 http://inperc.com/blog2/?p=570#comment-871 My understanding is that this isn't stereo vision, but something more interesting, using a time-of-flight camera chip developed by a small Israeli firm. The chip functions like radar or sonar, but emitting light (outside the visible slice of the spectrum) with a known phase. The light reflects off objects in the scene and some of it bounces back into a camera lense which focuses it onto a camera chip, in which each pixel measures both intensity AND phase. The difference in phase between emitted and sensed light depends on how long a path the light took between the source and the reflecting object - so the operating principle is like sonar or radar. Since phase is measured at each pixel, a depth value can be calculated for all pixels that are close enough to the light source and reflect a sufficient fraction of the light (i.e. not too black, not too shiny). The resulting depth maps are better than can be done with existing stereo systems (denser, and with an error that doesn't depend on distance), and also require little processing power to interpret - since distance is being directly measured, not inferred using image pair correlations. The time-measurement also allows for interesting tricks like peering through translucent media like fog or water, using time/distance to filter out the obscuring material to reveal a more substantial hidden surface. Many companies are already producing similar cameras, but none on the scale or price point of the Kinect. Interesting possibilities! My understanding is that this isn’t stereo vision, but something more interesting, using a time-of-flight camera chip developed by a small Israeli firm. The chip functions like radar or sonar, but emitting light (outside the visible slice of the spectrum) with a known phase.

The light reflects off objects in the scene and some of it bounces back into a camera lense which focuses it onto a camera chip, in which each pixel measures both intensity AND phase. The difference in phase between emitted and sensed light depends on how long a path the light took between the source and the reflecting object – so the operating principle is like sonar or radar. Since phase is measured at each pixel, a depth value can be calculated for all pixels that are close enough to the light source and reflect a sufficient fraction of the light (i.e. not too black, not too shiny).

The resulting depth maps are better than can be done with existing stereo systems (denser, and with an error that doesn’t depend on distance), and also require little processing power to interpret – since distance is being directly measured, not inferred using image pair correlations.

The time-measurement also allows for interesting tricks like peering through translucent media like fog or water, using time/distance to filter out the obscuring material to reveal a more substantial hidden surface.

Many companies are already producing similar cameras, but none on the scale or price point of the Kinect. Interesting possibilities!

]]> Comment on Topological data analysis by Roman Shapovalov http://inperc.com/blog2/2010/01/25/topological-data-analysis/comment-page-1/#comment-864 Roman Shapovalov Wed, 27 Jan 2010 21:13:46 +0000 http://inperc.com/blog2/?p=338#comment-864 Have you heard about Locally Linear Embedding? It does some kind of non-linear dimentionality reduction, though I'm not sure it can help to investigate topological properties. Hope it will help you! Have you heard about Locally Linear Embedding? It does some kind of non-linear dimentionality reduction, though I’m not sure it can help to investigate topological properties. Hope it will help you!

]]> Comment on No web apps, for now by Peter http://inperc.com/blog2/2009/11/27/no-web-apps-for-now/comment-page-1/#comment-862 Peter Sat, 28 Nov 2009 18:41:11 +0000 http://inperc.com/blog2/?p=250#comment-862 It’s a fair point. However, bio/med researchers are often concerned about analyzing the original images. Most likely, if someone comes with a bmp or tiff file and sees that it can’t be analyzed, he would simply walk away. I certainly wouldn’t try to force them to convert everything to png. A partial solution could be to have it done automatically before uploading but that would be more like a hybrid then a pure web app. -- Peter It’s a fair point. However, bio/med researchers are often concerned about analyzing the original images. Most likely, if someone comes with a bmp or tiff file and sees that it can’t be analyzed, he would simply walk away. I certainly wouldn’t try to force them to convert everything to png. A partial solution could be to have it done automatically before uploading but that would be more like a hybrid then a pure web app. — Peter

]]> Comment on No web apps, for now by overrider http://inperc.com/blog2/2009/11/27/no-web-apps-for-now/comment-page-1/#comment-861 overrider Fri, 27 Nov 2009 18:28:05 +0000 http://inperc.com/blog2/?p=250#comment-861 Why shoud one load 1000×1000 bmp file while PNG exists? Moreover, if sensors naturally save photos in JPEG, it is the option too. Why shoud one load 1000×1000 bmp file while PNG exists?
Moreover, if sensors naturally save photos in JPEG, it is the option too.

]]> Comment on A common view of digital imaging by uberVU - social comments http://inperc.com/blog2/2009/10/22/a-common-view-of-digital-imaging/comment-page-1/#comment-860 uberVU - social comments Mon, 26 Oct 2009 19:03:05 +0000 http://inperc.com/blog2/?p=223#comment-860 <strong>Social comments and analytics for this post...</strong> This post was mentioned on Twitter by lonifasiko: A common (and very far from reality) view of digital imaging http://bit.ly/3D9nLR... Social comments and analytics for this post…

This post was mentioned on Twitter by lonifasiko: A common (and very far from reality) view of digital imaging http://bit.ly/3D9nLR...

]]> Comment on Site updates by Math World | Site updates http://inperc.com/blog2/2009/09/13/site-updates/comment-page-1/#comment-859 Math World | Site updates Wed, 16 Sep 2009 01:57:04 +0000 http://inperc.com/blog2/?p=194#comment-859 [...] See more here:  Site updates [...] [...] See more here:  Site updates [...]

]]> Comment on Pixcavator 4.2 released by Peter http://inperc.com/blog2/2009/06/10/pixcavator-42-released/comment-page-1/#comment-858 Peter Mon, 06 Jul 2009 19:56:00 +0000 http://inperc.com/blog2/2009/06/10/pixcavator-42-released/#comment-858 Thanks for the compliments and the feedback. The first one must be a bug… It would help me a lot if you can describe the circumstances of this so that I can reproduce it. My email is saveliev@inperc.com. The second one isn’t a bug but I can’t say it’s a feature either. I just haven’t been able to come up with a better way of handling this dilemma: as I move a slider I definitely want both the contours and the summary statistics (and the table) to be updated instantly, but it’s just too slow… So I just imposed this limit – you are the only one who noticed so far! --Peter Thanks for the compliments and the feedback.

The first one must be a bug… It would help me a lot if you can describe the circumstances of this so that I can reproduce it. My email is saveliev@inperc.com.

The second one isn’t a bug but I can’t say it’s a feature either. I just haven’t been able to come up with a better way of handling this dilemma: as I move a slider I definitely want both the contours and the summary statistics (and the table) to be updated instantly, but it’s just too slow… So I just imposed this limit – you are the only one who noticed so far!

–Peter

]]>