by Roger N. Clark
Digital is Better than Film!
No It's Not! Film is BETTER than Digital!
No, You're wrong! No, you're wrong!
Guess what? Flat statements like this are both wrong, and both right!
Why? There is no one single correct answer. Thus, depending on what you want to do, one tool may be better than another.
It is often asked if digital cameras meet or exceed the imaging quality of film and what are the advantages and disadvantages. Let me first consider 35mm versus digital.
I've done digital imaging and image processing in the scientific world since about 1977, so I am very familiar with the technology and its use. I set high standards for myself in all my work and play. Currently I use both film and digital.
Here are some of the issues in the Digital vs Film debate:
More details: spatial resolution, dynamic range and signal-to-noise
So, how many megapixels do you need (Spatial Resolution)?
The above plot is a summary of my research. This plot, digital megapixel equivalent
versus film speed, refers to spatially resolved detail.
In the experiments I've done, using a variety of films and digital sensors,
my data and test results show quite a range of answers concerning whether
film is better than digital with regard to spatial resolution. Remember,
spatial resolution is only part of the story of perceived image quality.
Detail of this research can be seen at:
Film versus Digital Information,
An Image Detail Test: Scanned Fujichrome Velvia Versus 6-MPixel Digital,
Image Detail (How much detail can you capture and scan?), and other articles at:
Figure 1, above, shows the pink zone where film can give better results. If your digital camera works in the green zone, a quality digital camera can give better results. But that is not all, it really depends on WHAT you are trying to do.
In Figure 1 above, the derived digital megapixel equivalents are fuzzy numbers. For example, the Fujichrome Velvia 35mm equivalent is 10 to 16 megapixels. The 10 and 16 are soft. By soft I mean by several megapixels, like 10 meaning 8 to 12, and 16 meaning 14 to 18. You can see what I mean by trying to compare some of my test images yourself. The vertical bars for color represent the varying resolution one gets from Bayer Sensor digital cameras: color has less resolution than luminance. So the upper part of the bar represents the megapixel equivalence for luminance information, and the bottom gives the color information.
Examples: If you want to do landscapes and use the finest grained films, like Fujichrome Velvia as ISO 50, current top end cameras like the 6.3 megapixel Canon 10D will not deliver the detail of fine grained film. The 11 megapixel canon 1Ds comes close to 35mm Velvia. Example 2: If you do wildlife or action photography, you want to record the image with fast shutter speeds. You need fast film/digital sensor speed and or fast lenses. Some stop action wildlife photography requires shutter speeds faster than about 1/2000 second. Slow speed film will simply not work in many lighting situations. Professional wildlife photographers often use 100 speed film pushed one stop to ISO 200. In the above plot, we see that a 6 megapixel camera operating at ISO 200 will deliver more image detail than the corresponding ISO film. So, determine what film you have been using for what conditions. Check the above plot and see what digital megapixel equivalent that corresponds to and choose a digital camera that would be better.
If you want to see the details of my research, including image comparisons
for digital up to 200 megapixel equivalent and film, including 35mm,
medium and large format (up to 8x10), go to
my image detail main page
and read the other articles. Specific articles on film versus digital
camera spatial resolution include:
Image Detail (How much detail can you capture and scan?),
An Image Detail Test: Scanned Fujichrome Velvia Versus 6-MPixel Digital, and
How much to sample to record "all the detail?"
But spatial detail is only part of the story.
Dynamic Range and Signal-to-Noise
Figure 2. Example Transfer functions of one digital camera, one slide film and one print film.
There seems to be an urban legend that says digital cameras have less dynamic range than film. The legend is wrong. The above plot shows the comparison of a DSLR with print and slide film. The slide film records only about 5 photographic stops of information (a stop is a factor of 2, so 5 stops is 32). The print film shows about 7 stops of information. The digital camera shows at least 10 stops of information (this test was limited to 10 stops). Other tests show the Canon 1D Mark II camera has about 11.6 stops of information (a range of 3100 in intensity). Other DSLR cameras, like the Canon 10D have around 11 stops. Point and shoot digital cameras, somewhat less.
Examine the scatter in the data in the above figure. The scatter is due to
noise. The digital camera has several times less noise at medium and high
intensities, and dozens of times less noise at low intensities.
The low scatter of the blue points shows the low noise of the
digital camera, and it is lower at all scene intensities.
For further information, see:
Dynamic Range of an Image 2: Dynamic Range and Transfer Functions of Digital Images and Comparison to Film
Dynamic Range of an Image 1: How many bits do you need? (Intensity Detail of an Image)
The Signal-to-Noise of Digital Camera images and Comparison to Film.
Apparent Image Quality (AIQ)
While most who have worked with digital camera images agree that because of the "smoothness" of digital images, they can be enlarged more than film images. My testing shows that fine grained film has higher spatial resolution than 8-MPixel digital camera images, but the digital camera images have several times higher signal-to-noise. People infer image quality as a function of both spatial resolution and signal-to-noise. While this is a subjective concept, I've started some experiments to test this "Apparent Image Quality," or AIQ. My initial results (example references below) are showing to first order that there is an approximate equal trade for signal-to-noise ratio versus spatial resolution. Thus, if you had a digital camera that produced 8 megapixels and twice the signal to noise as fine grained film, the apparent digital camera megapixels could be doubled when comparing to film. So that 8-megapixel image may have the "apparent image quality" of 16 megapixels if compared to the lower signal-to-noise film. Since my tests show the spatial resolution of fine grained 35mm film like Fuji Velvia is around 16 MPixels digital equivalent, then that 8-MPixel digital camera probably produces similar "apparent image quality" to 35mm fine-grained film.
But high end DSLRs, like the Canon 1D Mark II have several times the signal-to-noise ratio of film, so this boosts the apparent image quality by the same factor as the ratio in the signal-to-noise values, propelling the 1D Mark II images higher in "perceived image quality" than fine grained 35mm film. While my research is preliminary, it does seem to agree with what people are saying, and because people look at different things (image smoothness versus spatial detail), it shows there is a lot of room for interpretation.
Note, too, that there are differences in signal-to-noise ratio between
different digital cameras, mainly due to their differing sensor and pixel
sizes. Larger pixels in larger sensors, in general, produce higher signal-to-noise
images because larger pixels collect more light from the lens.
Digital Cameras: Does Pixel Size Matter? Factors in Choosing a Digital Camera
I will continue testing on AIQ to further improve the AIQ concept. If my research trend holds, then the ~16 megapixel cameras with large sensors will have ~64 AIQ MPixel film equivalent, which is well into the higher medium format size range. Impressive! But small pixel size cameras, typical of "point and shoot" models, will probably have AIQ values several times less than the large pixel size DSLRs, even given the same number of megapixels in each camera.
My working AIQ equation that fits my data so far is:
AIQ = StoN18 * MPix / 20.0,
where StoN18 is the signal-to-noise of the sensor on an 18% gray target, assuming a 100% reflective target just saturates the sensor, and Mpix is the megapixel equivalent. It is clear from existing data that the trends predicted in the equation are correct, but there in uncertainty in the magnitude of the differences as there are only a few data points so far. So to say camera X is twice as good as camera Y is taking the data too far, but to say camera X is better than camera Y based on the AIQ is clearly within the bounds of the data when there is a large difference in the AIQ value.
Here are typical parameters (consider the AIQ as a relative scale whose magnitude is relative and the numbers are fuzzy to about 30%). For example, to say the top DSLR in the table (AIQ = 76) is better than Medium format (AIQ = 70) is taking the numbers too far.
Full Well Pixel signal-to-noise Millions Camera Examples Capacity Spacing 18% Target of Pixels (Electrons) (microns) (StoN18) (MPix) AIQ Digital MF* ISO 100 50,000 6 95 39 185 Digital SLR ISO 100 50,000 8 95 16 76 Digital SLR ISO 100 50,000 8 95 8 38 Digital SLR ISO 400 50,000 8 47 8 19 Digital P&S ISO 50 22,000 2.8 63 8 25 Digital P&S ISO 100 22,000 2.8 44 8 18 Digital P&S ISO 400 22,000 2.8 22 8 9 Digital P&S ISO 50 22,000 2.8 63 5 16 Digital P&S ISO 100 22,000 2.8 44 5 11 Digital P&S ISO 400 22,000 2.8 22 5 6 35mm Film: Film Fuji Velvia 50 - - 18 16 14 Film Fuji Provia 100 - - ~20 7 7 Medium format: 6x4.5cm Film Fuji Velvia 50 - - 18 50 45 Medium format: 6x7cm Film Fuji Velvia 50 - - 18 78 70 4x5 Film: Film Fuji Velvia 50 - - 18 240 220 8x10 Film: Film Fuji Velvia 50 - - 18 960 860
The above table shows Digital SLRs (DSLRs) with larger sensors have a clear advantage over digital point and shoot cameras, which are slightly above slow speed 35mm film. Note that the digital advantage over film is smaller for small sensor P&S cameras than for DSLRs with larger pixel sizes.
* The Digital MF is for a hypothetical medium format sensor. A 39 megapixel MF digital sensor has recently been announced, and if the sensor has full well capacities and sensor noise similar to existing Digital SLR cameras, it could have specifications like this.
The reason for the large increase in image quality with higher signal-to-noise ratio is that one can increase spatial resolution of an image by image restoration at the expense of signal-to-noise ratio. Here is one such example: Image Restoration Using Adaptive Richardson-Lucy Iteration. In this example, a high end DSLR images showed about a factor of 2 increase in spatial resolution, thus an effective increase of 4 times the megapixel count, from 8 to about 32. Here is another example with a factor of 3 increase in spatial resolution using averaged images to increase the signal-to-noise ratio: Saturn with a Telephoto Lens.
If you want to do landscapes with maximum image detail, for the "feeling of being there," try large format view camera photography.
A MAJOR CONSIDERATION WITH DIGITAL CAMERAS: SHUTTER LAG. This is the time it takes for the camera to acquire, track and lock focus and trip the shutter. Many (most, all?) point and shoot digital cameras seem horrible in this regard. Those I've tried seem an exercise in frustration, taking one to several seconds to take a picture. Such long delays make it difficult to get a candid picture of your kids at play, and make fast action like sports or wildlife pictures all but impossible.
Digital SLR cameras, like the Canon 1D Mark II, 10D, 20D, 1Ds, and Nikon D100, D200, D300, D3 have excellent (low) shutter lag and are equal to (or better than) 35mm film cameras concerning shutter lag.
The Achilles heel of Digital Cameras! Digital cameras, especially cameras with removable lenses have one major issue: DUST ON THE SENSOR. On such cameras, dust becomes very noticeable on images obtained with lens apertures of f/11 and smaller (higher number, like f/16, f/22). Thus for landscape or other photography requiring small apertures to maximize depth of field, film has the advantage. While you can clean the dust off, it is a short term solution, especially for long sessions in the field where it is impractical to clean the sensor multiple times each day. If you do not use small apertures, or only work indoors or clean environments, this may not be a problem.
I now use almost exclusively digital (DSLR) and occasionally film (4x5). I use Canon a 30D 8 megapixel camera and a Canon 1D Mark II 8-megapixel camera when I need fast speed with higher shutter speeds. For wildlife action photography, I feel can get better images with the digital than I can with 35mm film. For examples, see my Birds and Birds in Flight, Bears, or Africa galleries.
For landscapes, I use a 4x5 camera and Fujichrome Velvia film (more than 200 megapixel equivalent), then scan the film. See my Landscape Favorites Gallery for examples. If I do not have time to set up the 4x5, I'll use the DSLR at low ISO speed, and sometimes do multiple frames to mosaic a larger scene, effectively creating an image with more pixels. See: Large Digital Mosaics as a Substitute for Large Format Film for digital mosaics that surpass 4x5 film.
References and Other Reading
See Norman Koren's site, http://www.normankoren.com/Tutorials/MTF7.html were he concludes the Canon 10D 6-megapixel digital camera is equivalent to Fujichrome Provia 100F. Provia 100F plots a little higher on Figure 1 than Ektachrome, so Koren's testing regarding spatial resolution is in close agreement with mine.
This web site reaches similar conclusions: 8-megapixel DSLR cameras produce better (higher spatial resolution) images than 100 speed Fujichrome Astia film: http://www.wlcastleman.com/equip/reviews/film_ccd/index.htm.
Yet another film-digital comparison, this one showing an 11-megapixel digital camera produces clearly sharper images than Provia 100F film: http://www.photographical.net/canon_1ds_35mm.html.
The concept of signal-to-noise and its effect on apparent image quality is not new and is discussed at: http://www.normankoren.com/Tutorials/MTF7A.html#Shannon along with other references.
Another article about perceived image quality is at: http://www.luminous-landscape.com/tutorials/dq.shtml. Note when reading this that some of the signal-to-noise values are higher than possible from photon statistics alone.
An article about image restoration and the cost in noise from the Space Telescope Science Institute: http://www.stsci.edu/stsci/meetings/irw/proceedings/whiter_damped.dir/whiter_damped.html
Finally, here is a demonstration of enhancing spatial resolution, trading signal-to-noise: http://clarkvision.com/astro/saturn.03.02.2004 by acquiring 99 images of Saturn, adding them together, then improving spatial resolution with Richardson-Lucey adaptive restoration.
First Published, May, 2002.
Last updated February 9, 2016