Monday, October 28, 2013

Field Level at Gillette

Thanks to Tom Briere and his son Evan, eleven NVPC members had a fantastic opportunity to shoot at field level of the Gillette Stadium on Saturday, October 26th—up close and personal!  The UMass Minutemen vs Western Michigan Broncos football teams played a very challenging game right down to the wire, although UMass lost 31/30.  Winds were more than 20 mph, making it a very chilly afternoon.  Mike Los, a former alumni of UMass, also joined the band on the field.  

Field trips present club members an opportunity for camaraderie and sharing photo expertise.

A few shots from members are below.  Enjoy!

Thursday, October 24, 2013

What Does Digital Camera Noise Look Like

What Does Digital Camera Noise Look Like, Causes, and Processing Mitigation
“KNOW Thy Camera” - Every camera will experience noise in some way What does noise look like?
Color dots or speckles of various intensities (above and below the average intensity) in an image readily observed at 100% magnification on a computer especially in continuous tone areas and shadows especially when brightened
Causes of noise 1.
2. 3.
4. Types of Noise
2. 3.
Camera sensor :
Size of photo sites
o Noise is proportional to the size of the sensor at same capture conditions o Full Frame ~15 times larger than P&S and 4 times larger than 4/3rds
o Small sensors with larger number of pixels (MP) means more noise (e.g., in
4/3rds noise at 400 ISO will approximate the noise at 1600 ISO in full frame
  •   Camera circuitry
  •   Camera algorithms
    ISO amplification, i.e., higher than base ISO means more noise
    Under Exposure
    many pixels are captured with little tonal separation, i.e., low signal to noise ratio by not exposing to the right (ETTR). As tones become brighter (ETTR) noise is less with naturally higher signals due to more light being captured by the photo sites Long Exposure Times (over heating photo sites)
    Long exposure noise may start as soon as ~1/5 second in some cameras, and certainly when >30 sec with colorful speckles representing hot pixels which seem to have a pattern to them that is repeatable frame to frame at exposure conditions
    High ISO short exposure noise appears as random specks and vary from frame to frame Banding
    is very specific to the camera (appears as noisy horizontal or vertical banding especially in shadows and exacerbated with brightening)
Precluding/Minimizing High Noise Levels
In Camera
Purchase High End Larger sensor or newer technology (algorithms) cameras
... tradeoff – money (priceless)
Lower your ISO that may mean longer shutter exposures or opened up
apertures and perhaps using a tripod tradeoffs noise v motion v DOF Expose to the right (ETTR)tradeoffs none
Perhaps use some flash
Two Approaches to Mitigating Noise
Cameras have:
  1. Black frame subtractions for long exposure - a second exposure without
    the shutter opening for exactly the same length of time)
  2. Noise reduction algorithms in camera
Post Capture Tools focus their methods on: 1. Luminance
2. Color
3. Banding
Presented to NVPC 23Oct13
Demonstrations by Meyer Franklin and Jim Coutré
Page 1 of 2

Mitigating Image File Noise Levels
In Post Processing consider Global and/or Selective areas of mitigation application
  1. RAW development
    Light Room
    Adobe Camera Raw (ACR) in Photoshop
    Stand Alone Noise Reduction Software
    Stand alone complete processing suites from RAW through artistic final output having noise reduction features

  2. In Photoshop
  3. Plug-ins Dfine, Noise Ninja, DeNoise, etc.
  4. Blending multiple exposures in PS layers blending mode for the random high
    ISO/short exposure type noise
It is important to remember that:
  1. Noise tolerance is a personal decision
  2. Noise is less or not obvious and objectionable when the image view is small (Web posts and 4x6’s can take a lot of noise – thus smart phone and P&S cameras).
  3. Noise is more obvious and likely objectionable by 8x10 and larger
  4. Significant crops from the captured frame or excessive brightening exacerbates
    noise in final output
  5. Mitigation always reduces details and can make image look “plastic”
Suggested Work Flow Considerations
  1. Mitigate Noise preferably before sharpening, clarity, micro-contrast, detail
    enhancement, etc. since those adjustments actually exacerbate noise.
  2. Consider Selective use versus Global or combination of both for mitigation
  3. Consider using Smart Object layer in Photoshop
  4. Last step - Sharpen for output with Global and/or Selective techniques
  5. Exception for only LR rendering through final output by default LR and ACR use
    Adobe’s pre-sets for noise and sharpening, however their processing order is secret and should be okay but their mitigation is Global and arguably not yielding the best results possible
  6. When using other stand alone or plug-in mitigation software LR and ACR RAW rendering “automatically” add sharpening and noise mitigation unless you set it to “OFF” or “zero”
So, KNOW Thy Camera, your tolerance level in your final crop, and its intended final output size as you make your pre-shutter click parameter choices ... and then shoot RAW.
There are many references and tutorials available through web searches relative to noise and mitigation in digital files. Some reference urls:
Digital Exposure Techniques:
Digital; Camera Image Noise Part 1
Digital; Camera Image Noise Part 2
By Nasim Mansurov
Nik Software by Google Dfine is only one module
DeNoise by Topaz Labs
Noise Ninja by Picture Code Included in the newer Picture Ninja
Page 2 of 2
Presented to NVPC 23Oct13 Demonstrations by Meyer Franklin and Jim Coutré 

Addendum to My Over View Handout of Last Evening
Additional comments to the In-camera Mitigation of noise and a highly recommend tip to all.
Post the presentation one member spoke with me. I suggested some images were significantly under exposed by the appearance of the histogram having no data to the right of the mid-point. That is a major tip off for high levels of noise that could easily have been prevented.
The response was that the LCD looked over exposed. You cannot trust the post capture image view on your LCD for proper exposure judgments, plain and simple because ... that screen is not calibrated for brightness. In fact you can adjust the brightness so that you can see it in various levels of ambient light (i.e., increased brightness for bright sun light or very low brightness for night time ambient conditions). What should you do? ... Trust your histogram and expose to have data going all the way to the right, ETTR, with no major spikes of data going up the right hand side. This results in the lowest noise at the ISO used, i.e. highest signal to noise ratio. Use the Red, Green, Blue histograms (if you have them as a set up option) otherwise carefully look at the White composite histogram (not nearly as good). In the RGB set of histograms, the Red channel will usually be the first to get to the right hand edge. A minor spike on the right edge by your exposure choice may be recoverable.
Yes you may see blinkies in your LCD, a possible blown highlight indicator and certainly caused by specular highlights (a thin and very small height spike on the far right hand edge of your histogram). If you see a lot of large area blinkies (sky or reflections off of tree leaves) or if the blinkies are a main part of the object of your capture (for example on a person), then yep it is blown and you are too far to the right to capture those areas with detail, so lower the exposure a bit ... but use the histograms to get to the best ETTR please.
In the attempt to reduce the noise from an under exposed capture (per the histogram) with mitigation software you all saw the degrading of detail ... well that is because the file does not have much detail there to begin with to differentiate noise from good data, i.e. low signal to noise ratio.
Then the further response was, but I was on a tripod with the aperture as wide as I could tolerate with Depth of Field (DOF) and could not go longer on the exposure because of movement in the frame becoming a motion blur. Solution - increase the ISO since even at a higher ISO, if the capture is ETTR without blow outs of significance, then the noise (due to higher ISO) is more readily reduced without nearly as much loss of details.
Suggested capture Tip - If you want to get the best exposure under trying conditions (even rapid variance of light level) of: lowest ISO, desired depth of field, no motion blur (even with VR or IS), and highest quality image (low noise), then set your camera to Manual, select the shutter speed to freeze the motion in your subject, set your aperture to the depth of field you wish, and then use your cameras Auto ISO setting. That permits the camera to select its best lowest ISO for exposure. If there is a significant blow out per your histogram (due to exposure metering or a contrasty scene exceeding the dynamic range of your camera) then just dial in some negative exposure compensation, i.e. -ev. This works well with or without a tripod. Any remaining noise in the image file will be easier to clean up while preserving more detail because you ETTR than if you didnt allow ISO to increase and accepted a gap of no data from the right hand edge. And yes, some scenes simply exceed the dynamic range of your camera, so I would accept histogram bunching on the left with good data on the right or those scenes would require exposure compositing or HDR techniques for optimization. Of course, this is just in my opinion, so try it and let me know what you think. Jim Coutré
© J. E. Coutré Addendum to my NVPC Noise DUG Handout 21Oct13

Just to expand on Jim’s discussion regarding push the histogram to the right,  when he talks about the data in the histogram, the attached chart shows you how much data is in each segment of the histogram.  You can see why push to the right gives you the most data to work with and the least room for noise.