Seam Carving for Content-Aware Image Resizing
Thanks to a forum member 'teabore' for spotting this pretty amazing new resizing technique from Dr Ariel Shamir and Shai Avidan of the Efi Arazi School of Computer Science. Revealed SIGGRAPH this new method of image resizing looks for seams (not simple columns or rows) of pixels with the 'least energy' (least contrast / change in detail) both vertically and horizontally in the image and then uses this to enable resizing without losing important image content such as human subjects or other detail. This technique can be used for reducing and enlarging images as well as removing items from the image which are not wanted (by manually painting 'negative weight' over an area of the image). But less of my waffle just jump in and watch this video of the algorithm in use, I assure you it will make considerably more sense. (Purist photographers look away now).
Jeez. That's a REALLY novel idea. When you get right down to it, it's actually very easy to do, and won't eat much CPU power.. Just really clever. Like all great ideas, it makes you go "why didn't I think of that" after.
neat, but mostly useless. Resizing is used to maintain the original image while constraining the dimensions to fit another use. Cropping is used to alter the framing and composition of a "raw" image. The technique in the link makes fundamental changes to the images, destroying the framing and composition of the original and having no consideration for the aesthetic motivation behind cropping.
Removal of people or objects from a scene is already possible without relying on complicated algorithms and careful adjustment of parameters. I'm not sure what this new technique brings to the table that wasn't already there. Well, other than yet another method to hamfistedly alter an image.
I'd guess any real usage would be in image analysis rather than much aesthetic application.
Removal of people or objects from a scene is already possible without relying on complicated algorithms and careful adjustment of parameters. I'm not sure what this new technique brings to the table that wasn't already there. Well, other than yet another method to hamfistedly alter an image.
I'd guess any real usage would be in image analysis rather than much aesthetic application.
Originally Posted by Kremlin,Aug 22 2007, 03:04 PM
Jeez. That's a REALLY novel idea. When you get right down to it, it's actually very easy to do, and won't eat much CPU power.. Just really clever. Like all great ideas, it makes you go "why didn't I think of that" after.
Trending Topics
Originally Posted by GT_2003,Aug 22 2007, 11:46 PM
neat, but mostly useless. Resizing is used to maintain the original image while constraining the dimensions to fit another use. Cropping is used to alter the framing and composition of a "raw" image. The technique in the link makes fundamental changes to the images, destroying the framing and composition of the original and having no consideration for the aesthetic motivation behind cropping.
Removal of people or objects from a scene is already possible without relying on complicated algorithms and careful adjustment of parameters. I'm not sure what this new technique brings to the table that wasn't already there. Well, other than yet another method to hamfistedly alter an image.
I'd guess any real usage would be in image analysis rather than much aesthetic application.
Removal of people or objects from a scene is already possible without relying on complicated algorithms and careful adjustment of parameters. I'm not sure what this new technique brings to the table that wasn't already there. Well, other than yet another method to hamfistedly alter an image.
I'd guess any real usage would be in image analysis rather than much aesthetic application.
It is an interesting technique and I can see where it might be useful if it works in real time on a web page (as implied by the example at the start) but it's going to change the composition of an image, it's aesthetic, so not a lot of use to us photographers

It would also be interesting to see how well an image survives repeated resizing

Thread
Thread Starter
Forum
Replies
Last Post







