Tag: graphics

Computational photography

Another alteration of a camera’s field of view makes it possible to shoot a picture first and focus it later. Todor Georgiev has developed a lens that splits the scene that a camera captures into many separate images.

capturing 3d from an image, allowing for sharpening after the fact etc.

In the same way that the transition from film to digital is now taken for granted, the shift from cameras to networked devices with lenses should be obvious. While we’ve long obsessed over the size of the film and image sensors, today we mainly view photos on networked screens—often tiny ones, regardless of how the image was captured—and networked photography provides access to forms of data that go beyond pixels.

a lot of words spent without saying anything non-obvious. the author fails to talk about Computational photography which is far far more interesting than the lame vintage filters we have seen so far.
2022-02-05:

Supervised labeling

A probabilistic formulation for semantic image annotation and retrieval is proposed. Annotation and retrieval are posed as classification problems where each class is defined as the group of database images labeled with a common semantic label. It is shown that, by establishing this one-to-one correspondence between semantic labels and semantic classes, a minimum probability of error annotation and retrieval are feasible with algorithms that are 1) conceptually simple, 2) computationally efficient, and 3) do not require prior semantic segmentation of training images. In particular, images are represented as bags of localized feature vectors, a mixture density estimated for each image, and the mixtures associated with all images annotated with a common semantic label pooled into a density estimate for the corresponding semantic class. This pooling is justified by a multiple instance learning argument and performed efficiently with a hierarchical extension of expectation-maximization. The benefits of the supervised formulation over the more complex, and currently popular, joint modeling of semantic label and visual feature distributions are illustrated through theoretical arguments and extensive experiments. The supervised formulation is shown to achieve higher accuracy than various previously published methods at a fraction of their computational cost. Finally, the proposed method is shown to be fairly robust to parameter tuning.

this system can produce tags on par with humans for many types of images.

SIGGRAPH

Proceedings of ACM SIGGRAPH 2006

Future images are going to be far more dynamic objects than today in about the same way as text has gone from something fixed and stable to something that is endlessly recombined and automatically collated, summarized, analyzed and hyperlinked.

2007-08-14: SIGGRAPH 2007 Papers. candy store!
2013-07-25: SIGGRAPH is always fun, and this is no exception.

2013-10-28: Looks shopped. I can tell from the pixels.

2013-11-10: Always amazing.

2014-05-17: The biggest disappointment with SIGGRAPH is how little of it makes it into daily life.

2015-06-29: cool new toys

2016-03-21: way too many explody things is getting old

2017-05-03: Very nice results, from SIGGRAPH of course. Don’t get distracted by the not very good clothing animation.

2018-09-02: Best of SIGGRAPH 2018.

Comparametrics

image stitching and correction algorithms

A particular class of functional equations, called comparametric equations, is introduced as a basis for quantigraphic image processing. Comparametric equations are fundamental to the analysis and processing of multiple images differing only in exposure. The well–known “gamma correction” of an image is presented as a simple example of a comparametric equation, for which it is shown that the underlying quantigraphic function does not pass through the origin.

For this reason it is argued that exposure adjustment by gamma correction is inherently flawed, and alternatives are provided. These alternatives, when applied to a plurality of images that differ only in exposure, give rise to a new kind of processing in the “amplitude domain” (as opposed to the time domain or the frequency domain).

Slower and sharper

2 of my favorite diversions recently got a boost through new technologies: snowboarding and photography.

getting a grip
For skis, a network of electrodes embedded in each ski base will apply an electric field to the ski-ice or ski-snow interface. This low-frequency electric field will cause ice and snow to stick to the ski base, increasing friction and limiting the speed of the skier. If any skiers out there care to go faster, just increase the frequency. A high-frequency electric field applied at the ski base has an opposite effect as it melts snow and ice just enough to create the same thin, lubricating layer of water, but without the refreezing/sticking phenomenon.

depth of field

This is an image of inclined crayons from a traditional F/8 imaging system. The depth of field is less than 1 crayon width. The foreground and background are badly blurred due to misfocus.

After simple color and object independent image processing the final Wavefront Coded image is formed. This image is sharp and clear over the entire image. Compare to the stopped down image from the traditional system. Wavefront Coding allows a wide aperture system to deliver both light gathering power and a very large depth of field.

A Wavefront Coded system differs from a classical digital imaging system in 2 fundamental ways. First, the light traveling through a Wavefront Coded lens system does not focus on a specific focal plane. Because of a special surface that is placed in the lens system at the aperture stop, no points of the object are imaged as points on the focal plane. Rather, these points are uniformly blurred over an extended range about the focal plane. This situation is referred to as “encoding” the light passing through the lens system. The special Wavefront Coded surface in the lens system changes the ray paths such that each ray (except the axial ray) is deviated differently from the path that it would take in a classical, unaltered lens system and therefore they do not converge at the focal plane.

The second difference found in a Wavefront Coded system is that the image detected at the detector is not sharp and clear, as discussed above, and thus must be “decoded” by a subsequent digital filtering operation. The image from the digital detector is filtered to produce an image that is sharp and clear, but has non-classical properties such as a depth of field (or depth of focus) that is much greater then that produced by an unaltered classical lens system of the same f number.

gregorLinks 1.0

i just discovered a bunch of submissions in my blog queue that had been sitting there for a year. guess it’s time i start to blog them (3/28):

a very smart graphics guy
john c. hart has some pretty interesting papers online, such as on ray tracing on mass market GPUs.

promoting the third culture
The third culture consists of those scientists and other thinkers in the empirical world who, through their work and expository writing, are taking the place of the traditional intellectual in rendering visible the deeper meanings of our lives, redefining who and what we are. edge.org.

nomenclature
peter saint-andre maintains a guide to the nomenclature of philosophy