Banding that is sometimes visible in the output of Canon sensors can be an emotive topic. Some people say it is a show-stopper, while others claim it is only a problem if you ‘horribly and unrealistically abuse files’ created in a test environment. Neither is quite correct, but banding is most definitely real and can absolutely ruin a file’s potential. Thankfully, however there is a software solution in the form of Nik Dfine, which is a noise control program that has a banding removal feature. I’m going to work through the processing of a file to show you how effective it is. While I have not used a Leica M240 at high ISO, I understand banding can also be an issue above ISO 3200. Dfine will work the same magic here too.
So where does banding occur in normal usage? Simple: it occurs when you bring about a substantial increase in exposure of very dark areas of the file. A classic subject for encountering this phenomenon is shown below, with the original capture. Note that exposure was governed by highlight burnout in the area of and around the sun (in this case, clipping is restricted to the actual sun’s disc alone). This means that the shadows are very dark. Landscape and documentary photographers often tackle lighting like this and I did on a daily, if not weekly basis, when shooting my Afghanistan projects on film.
Note that all files below are 1500 pixels wide. Click to see full-size.
Lets imagine that we want to increase the shadow exposure to bring in some detail to the thistle stems rather than leave it as a dark mass merging with the distant mountains. In the untouched file above the shadows are registering at about 1.5% to 3%.
Below is a file that has received 1.7 stops of additional exposure, a shadow lift of 25 (Lightroom 5), and blacks are set at +25. I have done no further processing to the file, because this would only complicate matters. I am fairly confident everyone will agree that I have not abused or over processed this file. You can now just about make out what lies beyond the stems at ground level.
Now lets look at the shadows at 100%. Oh dear, what do we have here? Yes, it is clearly visible vertical banding. This would be visible on an A3 print for anyone who gives the image any more than a passing glance.
If banding is visible with very conservative processing, what about if we lift the shadows from +25% to +50%?
Is anyone still doubting the severity problem for real-world images, processed in a sensible manner? And by the way, the degradation inherent to smaller jpegs uploaded to this site masks the problem when compared to viewed on screen or in print. There, the problem is even worse… The above file is a hideous mass of banding and noise.
When shooting film, backlit scenes like this were easy peasy. As long as there was shadow detail, you could overexpose the highlights giving you a dense negative there, flash the paper at the print stage to get the highlights over the exposure threshold of the paper and then further burn them in if required. You’d have nice shadow detail, for rich, well-separated shadows. The above Canon 5DII files are not in the same galaxy as film in terms of exploitable dynamic range. I had a limited amount of frames to choose from, as I have only 200 old files from my UK desktop on this laptop, but I was still able to find a file that would show this problem. The 5D III is definitely better and had this shot been from that camera it would have shown perhaps half as severe a problem, but it would definitely still be a major problem. Conversely, a Sony sensor with 14 stops of dynamic range would not only have given a lot more detail in the shadows, but they could have been lifted without remotely breaking sweat. This is why I own a Sony A7R and A7!
So what do you do when you encounter this problem when shooting with your 11.7 stop Canon? Well, you have two main choices:
Click below for next page.