Tuesday 19 May 2015

Field Marks - False Fringes

In this slightly unorthodox posting I am going to use a mystery photo to explore the topic of false fringes in digital bird images.

For those of us here on the western edge of Europe a Caspian Reed Warbler (Acrocephalus scirpaceus fuscus) is a major identification challenge and potential reward.  We know these birds should be occurring here.  Among the potential ID features for this taxon, 1st winter birds in autumn often show a fine pale tip to the tail (reference for example Birding Frontiers, Autumn Challenge Series, Garner 2014).  Caspian Reed Warbler has yet to be recorded in Ireland, though there may have been a few potential candidates, particularly in late autumn.  It may well take DNA analysis to clinch the first but that is no reason to avoid pouring over a potential candidate.

When Killian Mullarney had a brief encounter with an interesting Reed Warbler in Co. Wexford in October 2013 he obtained an image of the bird in flight depicting a possible pale tail tip.  The question is, are these blurred looking tail tips really pale or merely an effect of the lighting?  Or, perhaps motion blur, aliasing or some other artefact?  Killian has kindly let me publish this image with my analysis below.

Photographed using a Canon EOS 7D with 300mm EF IS USM lens
4th October 2013, 2:38pm
Shutter Speed 1/1250th second
F-stop 4.5
ISO 640
Auto exposure mode
Auto white balance

A crop of the tail area reveals what would appear to be reasonably broad whitish tips to each tail feather, though the image is very soft and both the wings and tail appear unnaturally pale or washed out looking.  There are slightly peculiar brighter highlight areas within the pale fringe.  I believe this is a telltale feature of motion blur, which I will demonstrate below.

Lighting can most certainly produce pale patches, including the impression of pale fringes.  This is more likely to be the case if the light is behind the subject and translucency is at play.  It might also occur if for example an edge is slightly raised and just about catches the light.  Judging by the shading of the twigs the lighting in this case was coming from the right.  The colours of the mantle, rump and uppertail coverts are reasonably well exposed so the bird certainly isn't in full silhouette.  Humans are not very well equipped to judge the direction of ambient lighting (as discussed HERE).  Despite this however I think it is reasonably safe to assume that the sun in this case was roughly at right angles to the bird and the observer, There seems to be an element of reflective glare off the right wing flight feathers and to a lesser extent the tail.  But it doesn't appear that the brighter tips are due to this glare.  Note for instance that the trailing edge of the wing is very similarly lit but doesn't show this pattern.  Overall I think the lighting would not account for the impression of such a clean, whitish fringe, confined as it is to around the tips of the tail, in this consistent pattern we see.  So I think we can eliminate lighting as a cause of this fringe effect in this case.

Exposure by itself would not introduce unnatural fringing around features in an image.  Suboptimal exposure reduces contrast which in fact will subdue fringing.  Focusing on the bird alone one might be forgiven for thinking the image is unusually dull and poorly exposed.  However, looking at the leaves and twigs the exposure of the image isn't actually all that bad.

Defocus and Motion Blur
Our mystery photograph appears to have an measure of both defocus (i.e. lens out of focus) and motion blur.  Some of the leaves and twigs are sharp while others are out of focus and others appear to have been in motion.  The bird appears to be slightly out of focus and is also clearly in motion.  In order to analyse focus errors here I have created two composite images, one with pale fringes and another without.  I have applied artificial Gaussian and motion blur tools to these images.  While these are not natural conditions, the trade off is to allow for a finer control of the variables involved.

The first test below is for defocus.  As stated in the graphic, defocus alone will not produce artificial fringes and in fact the opposite is true.  Defocus reduces contrast and obliterates fringes and other fine detail.

The second test below is for motion blur.  One of the difficulties in trying to replicate motion blur is that the blurred pattern is directional.  The longer the exposure the greater the likelihood that the blur pattern will change direction.  With such a short shutter speed in this case (1/1250th of a second) the motion blur is likely to be along a straight vector but it is impossible to work out exactly what that vector would have been, and it might vary at different points on the bird.  For instance the bird may have been moving up and away within the image while in the same instant it's tail may have been in down stroke.  Meanwhile a movement of the photographer's hand could shift the entire image in another direction entirely.  So the overall motion vector may differ from one place to the next on our subject.  After having played around with different motion blur angles I settled on zero degrees, or motion along the horizontal plane for my test images below.

What is particularly striking about this experiment is that the fringe is dissolved in places yet it is interspersed with patches of highlight.  This pattern is consistent with the fringe pattern in our mystery image.  Another point of course is that, while the feathers without fringes have a blurred, ghost-like edge there is no pale fringe introduced by the motion blur.  So motion blur alone would not explain the presence of bright, whitish tips to the tail of our mystery bird.  On the other hand the motion blur pattern obtained using the test image containing a fringe looks very like our subject.

Aliasing and Sharpening Halos
These are both artefacts which are introduced by the processor as a consequence of sharpening an image.  Aliasing would produce the impression of fringing in very low resolution images.  This is not relevant to this image as the resolution is acceptable.  Sharpening halos occur due to the overuse of sharpening tools such as unsharp mask.  Again this is not at play in this instance.  Sharpening introduces contrast to an image and halos will be found around other objects in the image, not just our subject.  Note for instance that the tail has a fringe yet the nearby secondaries of the right wing have not.  In this case the image overall is quite low in contrast and there are no signs of edge halos.  Just to confirm this analysis I have taken the test image without a fringe and created a false fringe by drastically over-sharpening the image and thereby adding an edge halo.  The resulting over-sharpened image has excessively high contrast and looks unnatural,  For more on these phenomena see HERE.

Purple fringing is a term used to describe a variety of factors that can give rise to the impression of a false fringes around often contrasting edges in images.  As the name suggests they are generally coloured and or chromatic aberrations so do not appear white and therefore would not account for the effect in this sample image.

While it is not possible to be totally definitive I think the evidence points towards this bird having had pale tipped tail feathers in life.  The overall lighting in the image doesn't support translucency or localised light reflection as likely causes for the fringe pattern.  For instance the fringes are found on the tail but not on the nearby trailing edge to the right wing, which is similarly lit.  This inconsistency would also tend to rule out an artefact as the cause of the fringes.  Motion blur on the other hand could account for the odd character whereby the fringes have patches of brighter highlight interspersed throughout.

While this analysis may not ultimately confirm the identification of this bird, hopefully it presents a flavor of the mechanisms giving rise to false feather fringes in bird images and how we might analyse images in search of them.

No comments:

Post a Comment