Friday, 20 July 2012

Exercise: Highlight Clipping

The purpose of this exercise is to look at the effect of overexposing an area of the photograph which results in highlight clipping. This is a digital problem – ‘when the highlights go, they go suddenly, as if they’ve been cut or clipped’ (Freeman 2009, p34) and thus it’s important to understand why it happens, in what situation it might occur, and how to avoid it (both whilst taking the photo and if that’s not possible, then in post-processing afterwards).
We are advised in the notes to pay particular attention to the following issues relating to highlight clipping:
  • Complete loss of visual information
  • Visible break in the form of an edge between nearly-white and total white
  • A colour cast in a fringe along the clipped white highlight
  • The colour saturation

The Exercise:
A scene with a wide range of brightness was chosen. The image below shows the scene photographed with the highlights just below clipping point.
Highlights just clipped (on-camera)

I had my camera set on manual exposure, so then increased the exposure by one stop (changing shutter speed) for the next photo, and then took 3 photos with progressively lower exposure, resulting in a total of five images, as shown below. In addition, I took photos 1/2 stop over and under the mid-point image. I have chosen an outdoor situation and used a tripod in order to keep the framing constant for this exercise.
Comparison between images above and below the 'Highlights just clipped' image

When zooming in, the shot that is one stop over shows some areas that look ‘blown-out’, and the shot with highlights just clipped shows some bright areas that may show some clipping problems also. The three decrease in exposure images don’t’ appear to have problems but are generally very dark (and in fact look like they may show clipping at the other end of the spectrum.

I have then analysed the photos in Lightroom, the screenshot of 1 stop over shows the highlight clipping turned on (this is the only image with any highlight clipping according to Lightroom).
1 stop over showing clipping of highlights in Red (Lightroom)
Zooming in on this area, with and without highlight clipping displayed.
1 stop over showing clipping of highlights in Red (Lightroom)

1 stop over (Lightroom)

I would have to say that there is a complete loss of visual information in the centre of this screenshot, though perhaps not to the detriment of the whole image. There is no visible break clearly seen, nor a colour cast along a fringe observed. The screenshot below shows the same area covered when the photo is taken at 2 stops over.
2 stops over (Lightroom)
In this image, the histogram (top right) shows clear clipping of highlights, and there is considerable loss of visual information in the overexposed areas. I can't clearly identify the other problems, though this is not to say they are not present! 

I then opened the 1 stop over image in Photoshop RAW Converter and used the recovery slider to 50%. This has brought back some of the lost detail in the overblown areas, as seen in the two screenshots below. The image on the left has had recovery applied.
Recovery applied to left image (1 stop over)
Recovery applied to left image (1 stop over) zoom into highlight area
Freeman, M. (2009), Perfect Exposure. East Sussex: Ilex

Sunday, 15 July 2012

Exercise: Scene dynamic range

This project builds on the last one by suggesting we photograph a range of scenes with different levels of dynamic range. I have chosen to photograph a mini-series about ‘Growth’, in the gardening sense (keeping it close to home again!)

Growth Settings:
  1.  Seeds (with artificial light) 1/125, f/4
    •  Bean 20sec, f/4
    • Spade 3sec, f/4
  2. Sunny sky 1/750, f/4
    •  Sky 1/500, f/4
    • Bright white cloud 1/1000, f/4
    • Grey could 1/350, f/4
  3. Growing (small): (overcast, cloudy) 1/250, f/3.5
    • Dirt 1/45, f/3.5
    • Porch 1/1000, f/3.5
    • Leaves 1/350, f/3.5
  4. Growing (larger): (sunlit) 1/125, f/4
    • White cord 1/500, f/4
    • Dirt 1/20, f/4
  5. Cutting on bench – with knife for reflection: (sunlit) 1/1000, f/2.8
    • Blade 1/8000, f/4.5
    • Shadows 1/60, f/4.5

The dynamic ranges were calculated as follows:
  1. Seeds: approximately 3 stops
  2. Sunny sky: approximately 2.5 stops
  3. Growing (small): approximately 4.5 stops
  4. Growing (larger): approximately 3.5 stops
  5. Cutting on bench: approximately 7 stops

Growth Images:

Seeds ~ 3 stops
Sunny sky ~ 2.5 stops
Growing (small) ~4.5 stops
Growing (larger) ~3.5 stops
Cutting on bench ~7 stops
The dynamic ranges of these images are generally quite low (and ‘safe’), apart from the one I planned to have high dynamic range, by including the highly reflective object (the knife) and shooting into the reflection to maximise the dynamic range. According to Lightroom there is a tiny bit of highlight clipping on the edge of the knife, so my camera copes fine with 7 stops. The sunny sky image has the lowest dynamic range at 2.5 stops, but I would not consider this image to be 'flat' in the way the notes describe.

Exercise: Your Cameras Dynamic Range

The dynamic range of a scene is the number of stops between the brightest and darkest areas of the image (highlight and shadow). The dynamic range of the camera is the number of stops it can capture in an exposed frame.

Dynamic Range (8.0)
The 30D scored very well in our dynamic range test, essentially matching the performance of the Canon EOS 5D. We test dynamic range by photographing a calibrated test target, whose lightest area is more than 13 stops brighter than its darkest and analyzing the image with Imatest software. Imatest measures how many stops the image shows distinctly at various quality levels. We report the High Quality level, which includes only steps that the camera shows with less than 1/10 of a stop of noise, and Low Quality, which includes steps with up to a whole stop. Though the test reports the number of stops detected, it's important to note that the testing setup is designed to show the maximum range possible for a given camera and the results are meant to compare one camera or ISO setting with another. It is very unlikely that an image shot of a typical, natural scene would achieve these levels of dynamic range. 

Most cameras we have tested show a significant decline in dynamic range as ISO increases: the better the camera, the smaller the decline. The 30D has unusually steady performance into high ISO ranges. At Low Quality, its range at ISO 1000 is only one stop less than its range at ISO 100. At High Quality, it loses less than 1 1/2 stops over the same range. It's only at 1600 and 3200 that big drops in range occur.

If the dynamic range of the camera is less than that of the scene, then either the highlights will be blown out, or the shadow areas lost of detail. According to the text notes, we can use the camera’s highlight clipping warning to find the brightest highlight, and then look at the noise in the dark end of the tones. There is a point when it is no longer possible to determine between noise and detail in the shadow zone. This is subjective of course, which could be one reason why camera makers don’t publish dynamic ranges of cameras generally.

I found a scene with a very high dynamic range, as suggested by the exercise notes.  This scene is in my garden and includes a white reflective surface (trampoline signage) and the dark part of the image is the indoor view through the window, as suggested by the course notes. I turned off noise reduction in my camera and set the ISO to 100. I took a couple of photos until there was no highlight clipping of the white sign.

f/4, 1/30sec

However when I loaded this image to my computer, there is very clearly some highlight clipping according to Lightroom, so I am a bit confused by this difference.

Clipping areas highlighted by Lightroom software

I then set the metering mode of my camera to spot, and measured the aperture/shutter speed combinations of various parts of the image. These are labelled in the image below and show the range is between 1/8000 and 1/20 (with constant aperture f/4), which I calculate to be 9 stops (also shown in sketch below).

Labelled shutter speeds


IMG_2789 is the first image that I open in Photoshop that has values a little less than 255 in the white label. This image is exposed at 1/2000sec, which is ~6 stops less than the image I took with my camera first which suggested little or no clipping of the sign. (??)


Perhaps I failed to correctly analyse the clipping on my camera when I took these images (a month ago now, as I have not had time to write up these notes). I then go to the dark areas of this image (there are many, in fact, it appears heavily underexposed!) and try to bring them up to look at the detail and noise, but it all appears quite noise and it’s hard to determine which is which. I’m not sure how I’m then meant to read off the range between the two ends as suggested in the notes.
The image that I took has roughly a range of 9 stops. There is no shadow clipping at the low end, but a lot at the high end. That means that my camera cannot capture the 9 stops of range.  Although I cannot conclude the exact dynamic range of my camera, I have certainly, through this exercise, become much more aware of dynamic range in the camera. I generally actually choose to only shoot scenes with a limited dynamic range as I am aware of this problem, however, it’s been interesting to have blown-out areas and see the impact of this on the image. I may try to use this more in my photography to push the boundaries somewhat.

Tuesday, 10 July 2012

Exercise: Colour cast and white balance

This exercise is fairly straightforward, and its aim is to show us the difference in using various white balance options. I always shoot RAW now, so I can edit this in post-processing, but do find (particularly when shooting under artificial lighting) that it makes a difference to how the image displays on the camera and thus the 'feel' of the image, so it does pay to get the WB right for the scene, or at least close to the scene.

The Exercise:
I have undertaken this exercise in the backyard, with simple props to maintain some constency between the images to hopefully see clear results. I photographed a similar scene under the three different lighting conditions; Sunny, Cloudy & Open shade on a sunny day. For each, I used a tripod, and used each of the four white balance options; Auto, Sunlight, Shade and Cloudy. All images were taken using ISO 200, 50mm lens, f/3.5 with varying shutter speed for the different light conditions.

The Images:
I have grouped them according to the three lighting conditions:

Sunny Day:
Sunny day images, with varied WB options
Slight differences are observed in the colours on the car, and on the green grass in particular. In this case it's really just a matter of slight preference, but the sunlight white balance and auto white balance have both done a good job.

Cloudy Day:
Cloudy day images, with varied WB options
Again, there are only slight differences between the colours in the car and grass. In this case any of the images are acceptable, with a slight preference for the cloudy option.

Open Shade on Sunny Day:
Open shade on sunny day images, with varied WB options
None of these images are particularly good actually, which is a bit surprising! I had thought the shady option would do a better job that what it has in this case.

Second Part of Exercise:
This part of the exercise called for a mixed-lighting scene.

These indoor shots were taken at sunrise at Caloundra, with a lamp on inside. The light levels were roughly similar between outside and inside as recommended in the course notes.  ISO was set at 400, and I have used my Canon S95 for this exercise, and would expect the results from my 30D to be very similar.

Indoor view to outdoors (mixed lighting), with varied WB options

In this situation, if I had to choose between AWB, Sunlight and Tungsten, I think the AWB does the best job, though it is close with Tungsten. The Sunlight version is much too yellow inside.

The AWB does a reasonable job mostly, but in general I choose the correct ISO for the conditions, and shoot RAW anyway.