Bob Newman says that early critics of high pixel counts on digital cameras were misguided
In the early days of digital photography there was a sustained campaign by some parts of the web media against cameras with high pixel counts. It was proposed that high pixel-count cameras inevitably suffered from more noise, were more susceptible to camera shake and absolutely needed better lenses. Another widely promulgated belief was that high pixel-count cameras suffered from the effects of diffraction more than low pixel-count ones.
It turns out that most of this was based on simplistic models and a faulty understanding of science, and that in most general-purpose photography, higher pixel-count cameras will produce better results than ones with lower pixel counts.
However, they do so at the cost of lower frame rates, larger files and a slight penalty in extreme low light. Here, I’ll try to address some of the advantages and disadvantages of high pixel-count cameras.
First, the primary source of noise in an image is what is called ‘photon shot noise’. It is the noise caused by the structure of light itself. Photons arrive in a random pattern that we call ‘noise’. The more photons you have in an image, the more this random pattern evens out and the less apparent the noise is. Thus, the level of noise is primarily controlled by the number of photons collected (at any given exposure).
However, the signal, or image information, is sampled more frequently across the frame, leading to a higher signal ‘bandwidth’. One factor of noise is the higher its bandwidth, the higher is its power. This power can, however, be limited simply by restricting its bandwidth. In image terms, this restriction of bandwidth is simply affected by the acuity of the human eye. So long as you view an image so that the individual pixels are not visible, you have applied the required ‘low-pass filter’ to limit the bandwidth.
The question this rather naturally raises is, if we view an image in such a way that an increased pixel count is invisible, what benefit does it bring? To answer this, we can look at the illustration, which I have derived from one produced by Dr Hubert Nasse of Zeiss (lenspire.zeiss.com/en/measuring-lenses-objectively-part-2).
This shows the modulation transfer function (MTF) of the same lens fitted to three cameras: 12 million pixels (red line), 18 million pixels (green line) and 24 million pixels (blue line). The graph shows contrast against feature size (in ‘line pairs per picture height’).
The lower-resolution sensors are delivering less contrast from about 600 line pairs per picture height, which is about equivalent to high-definition television (about two million pixels). Thus, even viewed at two million pixels, the 24-million-pixel sensor will deliver a sharper looking image than the 12-million-pixel one.
The difference is as pronounced as it would be between images produced by low and high-quality lenses. Thus, one of the benefits of a high pixel count is that, in effect, it makes all your lenses better.
Bob Newman is currently Professor of Computer Science at the University of Wolverhampton. He has been working with the design and development of high-technology equipment for 35 years and two of his products have won innovation awards. Bob is also a camera nut and a keen amateur photographer