Can your camera see better than you?

Discussion in 'Mirrorless Digital Cameras' started by rayfraser, Nov 10, 2004.

  1. My 1.3 Meg camera can barely see at 20/40. By that I mean when I
    take a picture of a standard eye chart from 20 feet with a standard
    lens (no zoom or telephoto) the lowest line that is readable is just
    one down from the top. Some Humans (7 million cones and 120 million
    rods) can clearly see the 20/10 line. Can a 9 megabyte digital
    camera "see" the 20/10 line? With a standard lens this would imply a
    cropped chart size of approximately 120x180 pixels. Please post a
    photo if you can demonstrate a 20/10 capability without any zooming.
  2. I can't seem to edit my original question. Too many years dealing with kilobytes, megabytes, gigabytes and terabytes - what I meant to type was "9 megapixel". Also 120x180 probably should be closer to 200x260.
  3. Spearhead

    Spearhead Moderator Staff Member

    Why should a photograph replicate human eyesight? That's not what photography is about at all.
  4. I don't happen to have a standard eyechart handy at the moment, for some reason (let me look. . . . no there's not one under the sofa, either) but yesterday I took delivery of my new Canon Powershot Pro1 (8mp) and where my Olympus E-10 (4mp) fell about even with 35mm/Tri-X, I think I've moved up to 35mm/Pan-X, or maybe even a bit better.

    Jeff, maybe sharpness doesn't matter to you (how are things on the Lomo forum?), but for what I do with my camera, it's essential. There's more to photography that just what YOU do with it, believe it or not.
  5. Spearhead

    Spearhead Moderator Staff Member

    If people comment that my photos are "sharp", I look to see what I did wrong. Photography transcends any specific technical parameter. If your job is photographing eyecharts, then it matters. Otherwise, it's just more technobabble from technoweenies rather than anything from photographers. I have yet to meet a photographer who spent any time photographing, printing, or framing eyecharts.
  6. Yeah, I buy all that, but how about having good bokeh?
  7. Good question, but what do you mean by a "standard lens?"
  8. Hi Ray, Just a guess on my part, but I suspect that a 1.3 Mp camera might not have the best lens in the world. The quality of the lens you use can come into play more than the sheer number of megapixels in your camera. Best wishes . . .
  9. To test your camera just print chart at -, and take a picture from 20 feet away.

    If you want only soft pictures with no sharp focus anywhere, replace your lens with a pinhole! Requiring an eye chart at 20 feet to be in focus (only a tiny portion of image) takes nothing away from ability to create artistic blur.
  10. If you put the best lens in the world in front of a 16 pixel digital sensor, you are not going to get very detailed pictures!

    A good pair of (human) eyes can resolve X - I don't know what X is in absolute terms; you can talk about rods and cones etc., but our eyesight doesn't work in the way that lenses and digital sensors do.

    A good lens can resolve Y - I can measure Y in absolute terms, but I have no idea how that compares to a decent pair of (human) eyes (X). The amount of Y that gets translated into film information or digital sensor information depends on the density of that medium. In the example above, nearly all of Y gets discarded because the sensor is only able to record 16 pixels of information.

    A 1.3 megapixel camera simply doesn't have the ability to record all of the information that a decent lens can capture. I have an old (Fuji) 1.3 megapixel camera too; the fixed hyperfocal lens is probably ok (similar quality to a decent disposable 35mm camera probably), but the sensor is lousy.

    A decent 6-8 megapixel DSLR is capable of recording most (if not all) of the useful information from a decent lens. I seen some posts here arguing that Canon's latest CMOS sensor (16.7 megapixels) is capable of recording more information than a decent lens can deliver.

    Now, whether that is as good as or better than a decent pair of (human) eyes is anyone's guess. However, depending on your view of the whole film v digital carnival, it's as good as any other (35mm) film camera with a decent lens. If an 6-8 megapixel DSLR sensor cannot resolve see your 20/10 line, then I doubt that a 35mm sized piece of colour print/slide film can too.
  11. The question is very interesting (academically speaking), but it does not seem to me very easy to answer. Intuitively, I would say that it would need a very good digital camera (or film camera) and lens in order to reach the resolution of the human eye. I suppose that when you mention standard lens is 50mm. Now, the human eye has a resolution of 1 cm at a distance of 20m. I think either way that the resolution of the human eye is given in the angular difference that two points can be resolved. A 50mm lens has angle of view 46 deg (I think I am right if not...) so at a distance of 10m takes an image of 17m (horizontally). This means that in order for the camera to resolve 1cm that the human eye resolves at this distance it has to see the difference of two lines that are apart for 1cm (this means 2 pixels per cm). So if I am not wrong the camera should have (horizontally) 1700X2=3400 pixels. Ans vertically we can say (due to the 35mm format) approximately 2000. So we need a 6.8MP camera but with real pixels not red blue green together, because we do not want the interpolation and a good lens to match the resolution of the camera. Now I am certain that maybe I have done a mistake but I liked the question.
  12. Excellant analysis Vasilis!

    The 35mm camera with a 50mm lens was (intentional or otherwise) designed to closely approximate our own vision when it comes to field of view. So my question may have been better phrased as: can a camera using a view angle of 46 degrees resolve better than a human eye? This properly restricts lens and imager size choices.

    I personally believe digital photography is on the verge of surpassing a human eye. Who will be first to post proof in the form of a photographed eye chart?
  13. Spearhead

    Spearhead Moderator Staff Member

    You can keep going on this, but it's ridiculous.

    The eye is nothing like a camera. The eye has a brain behind it. When I do an eye chart, I guess at what the letters are based on how close I think I am. I can get to 20/15 or something like that, but I know that I'm studying the letters figuring out whether it's E or F.

    The camera does nothing like this. It's a simple reproductive chain that doesn't allow for guessing.

    And this really has nothing to do with photography.
  14. Really I'm with Jeff S. on this kind of thing, but for the sake of discussion, if we get this eye chart photograph, photographed by this human eye surpassing digicam your talking about, and assuming we must look at it with our human eyes, that have been surpassed, uh.. how would we know it?

    Must be another faith based initiative.
  15. I don't get it. Cameras and optical systems have long been able to surpass the human eye in various ways - e.g. microscopy, spy cameras, x ray cameras, etc. And the human visual system has never been surpassed in other ways - e.g. white balance, feature extraction, object recognition etc. So what? What follows from that, or the eye chart thing? Beats me.
  16. This is a silly discussion.

    If we are talking about resolving power of digital cameras (vs. the human eye) then there are tons of posts (and web pages) about photographing USAF resolution charts, line pairs per mm, and all the rest of it.

    If we are talking about whether a camera can see something that my unaided eye cannot, then the the answer is duh! ever used a telephoto lens? :)
  17. Why add a silly comment to a discussion you consider silly?

    I specifically excluded the use of telephoto lenses "duh"! Did you read all of the posted comments before adding yours? If we allow other than a 46 degree angle for a lens then we could easily use binoculars as a legitimate substitute for comparing with human eye resolution.

    Are you suggesting using an USAF chart to measure a Human eye?

    Throughout the history of photography many efforts have been made to emulate the Human eye. I would even go as far to suggest that a good use of blur matches what an eye would perceive when staring focused on a single object in a scene, or even in highly emotional states what an unfocused eye has on its retina.

    I am dismayed that an easily performed test has been out right rejected by many members without even considering a possible academic knowledge increase. Digital cameras are just now gaining more pixels than cones in the human eye. Do you think there is no connection between bayer imager masks having more green pixels and higher sensitivity in the human eye to green?
  18. If 1.3 MP (with a good enough lens) can "barely see" 20/40, then doubling the pixel count horizontally & vertically would allow "barely seeing" 20/20. That would be 5.2 MP. So a 6 MP camera with a good lens should do it.

    As a sanity check, I printed out eye chart listed above. The 20/20 characters are 5/16 inch tall and 1/4 inch wide. At 20 feet, a 50 mm lens EQUIVALENT (a 31 mm lens on my Digital Rebel) will capture 14.2 feet x 9.6 feet at 20 feet distance (sorry, Vasilis, your math was wrong). Canon's 6.3 megapixels = 3072 x 2048 (Nikon's is about 3000x2000). So a letter on the 20/20 line of the eye chart gets about 5.5 pixels high x 4.5 pixels wide to cover it.

    I bet that's as readable as the 20/20 line is for most of us at 20 feet.

    And yes, excellent lenses are good enough for that. Just check the Digital Rebel's review at and it's obvious.
  19. Thanks a lot for the discussion. Unluckily I haven't got any idea of the eyechart you are talking about, according to my sometimes rather poor eyevision (enuogh for cars but not for a truck-driving license) and your calculations. I suppose my DSLR sees better than me, but this doen't stop me dreaming of a better one.
  20. Leonard Richmond - can you please post your 5.5 x 4.5 pixel letters of the 20/20 line of the eye chart... Will be good to see it, if possible.

    PS: It started as such a good question. I thought to myself - can someone puh-leeeese shoot and post that chart!.. Billions of digital pictures a day and yet no-one dares to shoot a piece of paper to prove a philosophical point.

  21. Ray mentioned the eye chart in one of his followup posts:

    "To test your camera just print chart at - ,

    and take a picture from 20 feet away."
  22. whats this got to do with photography again?
  23. O.K. I did it. 35mm lens wide open on *istD ISO 200 distance at least 6.3m. I hope I'll be able to upload.
  24. The chart is made in a nasty, quite uncommon font. I'd say we are quite similar, the Pentax and me.
  25. "So we need a 6.8MP camera but with real pixels not red blue green together, because we do not want the interpolation"

    Why not? That's how the human eye works to.
  26. Thanks for the charts. I think the answer to the original question is: "No, current [consumer] cameras can't see better than humans".

  27. Wow, thanks Jochen. I took your image, doubled its size so that I wasn't limited by my monitor's pixels, then sharpened it. The 20/15 line is clearly readable. The 20/13 line is almost - the last letter O could be a C. The 20/10 line is almost - the last letter D can't be determined. Looking around on the internet, I think that qualifies for 20/10 "vision".

    Of course, this was an underexposed jpeg. Would be nice to see what a properly exposed (I assume the paper was white) TIFF would blow up to be.

    And you said "35mm lens wide open". The lens gets sharper if you close the aperture 2 or 3 stops. If the image gets any sharper, it will certainly be better than I can see.
  28. O.K. about the pic I posted: Well stopping down would be nice, but according to the 20ft distance and the limited power of built in flashes... I had to do it this way. - No I didn't mess up my workflow, I shot RAW, converted to TIF, and saved as big JPG for posting. I doubt I got many mistakes in. The Pentax 35mm f2.0 should be a quite good lens. We are looking at single pixels here so I doubt stopping down would gain significantly more information. By the way the 35mm is a rather long standart lens on my 6.something MP chip, so the math wizards are right, our actual consumer DSLRs see less than helthy eyes.
  29. I prefer to use manual focus on my D100. But I wear eyeglasses.

    I bought a real rubber eyecup for the camera and sometimes I take off the glasses, compose in the viewfinder and let the camera autofocus do it's work and in that case my camera does see better than these old eyes.
  30. "35mm is a rather long standart lens on my 6.something MP chip,"
    The Pentax *istD sensor is 23.5 x 15.7 mm. So the mulitiplication factor is 1.53. So the 35mm lens is equivalent to 53.55mm in full frame (35mm film) format, about 7% longer than the "standard" 50mm lens (for full frame).
    That's close enough - the picture wouldn't look much different.
    20/20 is the average corrected vision. 20/15 is sharper than average. 20/10 is about the best that even the sharpest eyes can see.
    So the majority of people, even wearing corrective contacts or glasses, cannot see quite as sharp as your 6 MP digital camera with a "normal" lens. The best eyes may be ever so slightly sharper.
    Of course, your eye's resolution is much less away from the center of vision. Your camera/lens is MUCH sharper away from the center of the image than your eye is.
  31. One can't compare a human eye vs a still camera. Your eye is constantly moving collecting more details about the subject. A proper comparison is a human eye vs a video camera. Because of a rapid succession of frames, a moving picture device (eye, video camera) doesn't require all that high resolution that a still picture needs.
  32. No matter how much you move your eye, I'll bet you can't see the 20/10 line any better. Even while driving at high speed, you still benefit by having 6-7 million cones giving you an ability to read signs. While your brain depends on some minor eye movement even while staring, an ability to read a distant sign is primarily dependent on number and sensitivity of light receptors. Also many digital camera's can take video - or at least several frames per second. The Canon PowerShot Pro1 can shoot 2.5 frames per second, but any advantage of merging information from multiple shots would diminish after just 2-3 frames and only 1 frame may be needed to clearly identify the 20/10 line.

    I will concede I currently trust my 2 human eyes more than any manufactured sensor, but how many of us will be willing to allow electronic decisions by our cars based on cmos imagers?
  33. You cannot easily replace human eyes to a camera for a variety of reasons. One of these, is that the human eyes (and most eyes in the animal kingdom) are logarithmic sensors. That is why your camera (which is linear) cannot see as good as you in bright sunlit scenes. But I have a question, is there anything like a logarithmic sensor or film?
  34. For a 500 times extended dynamic range look at SMaL Technology For an amazing hacking effort on $20 dollar "Pure Digital" single use cameras which use SMaL imagers check out linux-hacker cameras.

Share This Page