Talk:Purple fringing

Analysis about possible effect of infrared needed

Although ultraviolet may be the cause of purple fringing, the reason of ruling out infrared should be justified better. Purple fringing is often seen also around specular reflections of sun, where infrared spectrum may be dominant. It is also easy to test with any infrared remote controller that infrared causes visible colors into photo. Attached GIF animation shows how infrared LED is seen by Canon 7D through EF 400 mm f/5.6 L USM lens, and how the infrared light at back expands into purple over the tree branch in front of the LED (if the GIF doesn't appear, click the frame).

Another GIF is created with a strong ultraviolet LED with a little bit of visible violet so that it lights up background of a tree branch, again. There is also normal daylight coming from windows. Unlike in IR-photo, the outermost fringing appears to be sharp and violet and changes into blurry blue towards center. Also, blue haze can be noted to spread over the whole branch.

Infrared LED as seen by Canon 7D camera thru EF 400 mm f/5.6 L USM lens. A branch of a tree is to demonstrate how infrared from back of an object blurs over and could be cause of purple fringing.

Effect of adding strong ultraviolet back-light with little bit of visible violet, with the same camera and lens. Note that the branch at front is not lit with ultraviolet directly, but all of the UV light comes from the background.

--Jarmniku (talk) 18:30, 28 July 2020 (UTC)[reply]

Better example photos

There are some nice cc-licensed photos showing purple fringing and partial correction by a UV filter here:

http://www.flickr.com/photos/steevithak/6128164049
http://www.flickr.com/photos/steevithak/6128164045 — Preceding unsigned comment added by 64.81.113.171 (talk) 18:58, 19 October 2011 (UTC)[reply]

The above linked photos are mine and you're welcome to use them in the article if they'd be helpful. Steevithak (talk) 23:12, 20 June 2013 (UTC)[reply]

Microlens explanation

I'm thinking of adding the following explanation, but it could do with some improvement:

The proposed explanation based on microlenses goes as follows. Each CCD cell in the photodetector is covered with its own microlens. The microlens pulls more light onto the tiny sensor area, and also effectively reduces what would otherwise be a gap between the cells. The microlens is far from optically perfect, in particular for off-axis light there can be reflection (scattering) at a shallow angle such that the reflected light is detectable by nearby cells. This causes blooming at high-contrast edges because light from the brightly-lit cells is scattered and picked up by nearby cells which are less brightly illuminated (the scattered light is also of course picked up by nearby brightly-lit cells but it makes little or no visible difference to those).

The characteristic purple colour of the fringes arises because there are three different colours of cells, namely cells detecting red, blue and green light. Each cell is comprised of microlens, coloured filter and photodetector. In the standard Bayer colour filter mosaic array, there are as many green-detecting cells as there are red and blue combined. The lenses over the green cells scatter purple light, therefore a bright or overexposed white highlight scatters more purple light than any other colour; that scattered light is picked up by the "wrong" sensors nearby. In unusual circumstances (for example, strongly coloured highlights, or purple fringes combined with chromatic aberration in the main lens or elsewhere) the fringes seen can be of other colours than purple.

Purple fringing can be reduced using narrow (slow) lenses and narrow apertures, and generally by composition (avoiding bright and over-exposed areas with high-contrast edges). Adjusting the lens geometry through lengthening the respective distances between aperature, lens and sensor will also help to cut-down the off-axis light entering the sensor, although the ideal setting will vary from camera to camera and lens to lens. The fact that lens geometry has an effect on purple fringing causes some people to attribute the problem to the lens rather than the CCDs, when in fact it is caused by the CCD microlens array.

In principle a camera with a larger CCD sensor will have less purple fringing (both because the shape of the microlens is less extreme, so less scattering, and because the scattered light will cover fewer nearby cells because cells are more widely spaced). Different (non-Bayer) CCD designs may have different fringing effects or no purple fringes.

Link to article about microlens arrays and filters (in the context of digital microscopy rather than photography, but the principles are the same): http://micro.magnet.fsu.edu/primer/digitalimaging/cmosimagesensors.html

Link to image produced by Fujifilm showing a side view of the microlens array and filters (in the SuperCCD configuration which is similar to Bayer) http://www.dpreview.com/news/0301/fujisuperccd/srstructure.jpg

'CMcD' 19 September 2006

I dispute the assertion that the microlenses over the green-sensitive photosites "scatter purple light". The non-green components of the light incident upon a green-sensitive photosite are absorbed by the green filter located below the microlens: they are not reflected back from it. Light reflected and "scattered" by a given microlens is not dependent on the colour of the filter beneath it. This seems very much like OR (and flawed OR at that) to me and should not be included in the article. 94.197.191.79 (talk) 01:22, 30 April 2009 (UTC)[reply]

I would dispute it, too, if it was in the article unsourced; but it's not, so it's not an issue. Dicklyon (talk) 03:29, 30 April 2009 (UTC)[reply]

I've edited article. "Image leak" is the same as blooming and it only allpies to CCD sensors, right? Anyway, it's cosmetic. More important is quoting dpreview.com. They say that purple fringing is a result of chromatic aberration at microlenses. That's obviously incorrect.
- The strongest refraction in normal cases is for violet light. Purple would mean that green light is refracted the most.
- Refraction dispersion is in range of single degrees. http://www.lumerical.com/..._x_normal_0.mpg -> few degrees won't matter with such a wide beam.
- On sample picture from dpreview we see severe fringing, it can't be explained by local (pixel-size) scattering. Anyway it should be anisotropic, in dpreview example fringes are radial from image center.
- As written above, filter is absorbing. Dielectric interference filter would change its transmission spectrum with changing incidence angle.
This discussion was made in many places on the web, also by me (I'm a physicist, optics major, researcher in Fiber Optics Group in Wroclaw University of Technology) and other specialists in optics. Conclusions was gathered by me and optoelectronics major on Polish Optical Forum (http://forum.optyczne.pl/viewtopic.php?t=9258).
It's hard to find logical arguments showing any influence of microlenses on chromatic artifacts in picture. Mostly they are a result of objective itself or light back-reflected from sensor to the lens and backwards.--palindrom (pl.wikipedia.org), 10:33, 27 December 2009 —Preceding unsigned comment added by 83.28.214.52 (talk)

Reference quality

The Pictureline article is held by many in forum discussions I've seen to be not the highest-quality reference ... basically its "other causes" of purple fringing are all sensor effects, so I listed them as such. OTOH, it's a commonly-linked reference and one of the top Google hits on "purple fringing" - David Gerard 14:31, 10 April 2006 (UTC)[reply]

Spectacular?

(Under "Mitigations")

Shouldn't that be "specular"? — Preceding unsigned comment added by Reffik (talk • contribs) 22:54, 30 January 2013 (UTC)[reply]

Reverted. Dicklyon (talk) 06:56, 31 January 2013 (UTC)[reply]