Mas existe uma alternativa criada pelos pesquisadores israelitas Shai Avidan e Ariel Shamir, chamada Seam Carving ou Retargeting. Trata-se de um algoritmo que redimensiona a imagem, mas detectando nela os pixels de menor importância, pra serem deletados. Uma vez processada, a foto pode ser espremida ou esticada, baseado da hierarquia de importância dos pixels gerada pelo sistema.

Bom, os próprios caras explicam a ciência muito melhor do que eu nesse vídeo :

Como isso ainda está em fase de testes e prototipagem, não existe nenhum programa comercial que faça todas as funcionalidades descritas. O que existe, pra experimentar, é um aplicativo de web chamado Rsizr. Apesar de não oferecer todas as funcionalidades como detecção facial ou diferentes interpretações para a imagem, dá pra experimentar e ver como a idéia funciona.

Agora eu só vejo isso virando um plugin pra Photoshop, ou até mesmo um filtro já incorporado em uma versão futura do programa. Pra quem mexe com editorial, e precisa de versatilidade na hora de layoutar, pode ser uma ferramenta bem útil. Imagino também a introdução de novos formatos de arquivo de imagem, com a informação da hierarquia de pixels já incorporada, pra que o redimensionamento possa ser feito em tempo real, o que seria útil pra adaptações de resolução na web e dispositivos móveis.

"content-aware resizing" is a resizing technique developed by Shai Avidan and Ariel Shamir using which Image resizing is done with an awareness to the content inside the image. Watch the below video for a demonstartion.

How this happens ? This technique is based on a concept of seam carving. Seam carving can be easily understood as 'Resizing image by gracefully adding or removing pixels from the not so important parts of image, thus preserving the contents of the image'.
Then 'How to determine those not so important parts of image?'. The answer as told by Shai Avidan and Ariel Shamir is

Seam carving uses an energy function defining the importance of pixels. A seam is a connected path of low energy pixels crossing the image from top to bottom, or from left to right. By successively removing or inserting seams we can reduce, as well as enlarge, the size of an image in both directions. For image reduction, seam selection ensures that while preserving the image structure, we remove more of the low energy pixels and fewer of the high energy ones. For image enlarging, the order of seam insertion ensures a balance between the original image content and the artificially inserted pixels. These operators produce, in effect, a content-aware resizing of images.

A detailed paper presented by Avidan and Shamir can be found at here.

This algorithm is already implemented in many languages and also used as plug-ins for GIMP and Photoshop.
You can find the GIMP plug-in here.

A dedicated group on Flickr for this kind of resize images can be found here.

But photo editing is changing once again. I came across a couple innovations in photo editing that show that efforts are being channeled towards what can be called “intelligent imaging”. To edit a photo today, you still have to understand the image: you and you alone know what is an object, what is a face, or where this object can (or cannot) realistically be; you and you alone can determine from where the photo should be taken, at what angle, with the focus on which object, etc. “Intelligent imaging” goes beyond that.

Better than a hundred words, two examples should do it:

- The first one is a video of conference by Dave Story, vice president of interactive design at Adobe. The website is in French, but the video at the end (in English) is the interesting part.

- The second one is a technology called “seam carving” (thanks Steve for the pointer), which is being developed in several universities and companies at this moment. The following video explains pretty much most of what there is to know. You can then download an open-source GUI software here if you want to try it out. For other examples and other software (including a Gimp plug-in), you can visit this Flickr group.

rsizr ist eine Flash-Applikation zum Resizen, Croppen und Drehen von Bildern. Zum Resizen wird der relativ neu veröffentlichte "Seam Carving" Algorithmus benutzt - sehr geile Umsetzung!

Infos zu "Seam Carving" hier, bzw. hier (pdf download, ~20MB).

First, a quick summary of the technique: Typically, when you resize a photo, you
just make everything in it smaller. Simple, right? Well seam carving is like a
smarter version of resizing. Your computer analyzes the photo and figures out what's
important and what isn't, then it eliminates the less important areas first as you
shrink the image.

That means you can resize a photo to be half as wide without shrinking everyone in
it, or even make a photo wider without making it look stretched. You can even pick
areas you definitely want to leave alone, or areas you definitely want to lose,
making it a super easy way to remove an ex or a stranger from a otherwise lovely
photo.

Well, enough already. You really need to give it a try for yourself.

Rsizr Free seam-carving image resizing on the web

A demonstration video

En effet, peu après avoir mis en ligne la vidéo, une version bêta du logiciel a été mise en ligne sur la toile, pour notre plus grand bonheur.

Je n'ai pas encore testé ce bijou, mais Bastien, blogueur @ plopblog.com, a présenté une démo qui vaut le détour, en partant de cette image :

Vous verrez, le résultat est impressionnant, surtout quand on le compare à un redimentionnement automatique...

(Via PlopBlog)

Edit

Je viens de tomber sur un post @ Emob qui parle du Seam Carving et qui donne quelques liens intéressants dont Rsizr, un site en flash qui propose de retailler vos images intelligemment directement en ligne.