September 18, 2014, 01:58:35 PM

Author Topic: Sensor Pixel Density  (Read 3999 times)

Mega

  • Guest
Sensor Pixel Density
« on: September 21, 2011, 02:12:30 PM »
Thanks to you guys, I have a much better understanding of why a bigger sensor is better, and also why we don’t see super-zoom cameras with large sensors (because the lenses would have to be impractically large).

However, I have another sensor-related question or two:

I see a lot of complaining about the megapixel count being too high on some of the small-sensor cameras.  I can see why, as I have downloaded full resolution samples from say a 16 MP super-zoom, and the image is quite fuzzy at full resolution.  But the same picture looks quite good at my monitor’s resolution (1920 x 1080) and even zooming in a few levels.

Question 1:  If I were to set the resolution to a lower setting in the camera, would that solve the issue?  In other words would setting a 16MP camera to 10MP output produce the same image quality of a camera with a 10MP sensor assuming all other variables were the same? (yes, this is hypothetical since there probably are not two matching cameras with only the sensor pixel density being different)

Question 2: If the answer above is "no", would it be better or worse to reduce the image size in software on the computer?  (I know jpeg is destructive compression, so assuming you are not working with RAW images, I would think reducing the size on the computer would be worse since it would be going through Jpeg compression twice)

Thanks in advance for your insight.

canon rumors FORUM

Sensor Pixel Density
« on: September 21, 2011, 02:12:30 PM »

K-amps

  • 1D X
  • *******
  • Posts: 1516
  • Whatever looks great !
    • View Profile
Re: Sensor Pixel Density
« Reply #1 on: September 21, 2011, 02:43:15 PM »
Q1: I would think no, because native pixels are sharper (more clarity) than interpolated pixels. This is assuming your Camera uses the whole sensor and not a crop at 10mp. I am saying this because if you set your monitor to a smaller resolution than it’s native resolution, it is interpolating pixels and it gets quite blurry. At magnification levels that are square roots or similar proportional numbers, the perceived sharpness is better e.g. 50% magnification looks better than 66% magnification on screen. So there is a relationship between the nativity vs. interpolation of the pixels. So I think the answer to this question might be a “No”

Q2) Might better to use a good software to reduce it. For example bicubic scaling is better for reductions as it naturally sharpens, and Bilinear is better for enlargements as it smoothens jaggies. Fractals are even better at enlarging. My guess is specialized software for resizing should out-do the in camera resizing.

Here's some info on the reduction/expansion algorithms: http://www.codinghorror.com/blog/2007/07/better-image-resizing.html
EOS-5D Mk.iii 
Sigma 24-105mm F4 ART; EF 70-200 F/2.8L Mk.II; EF 85mm L F/1.2 Mk. II; EF 100mm L F/2.8 IS Macro, 50mm F/1.8ii;  TC's 2x Mk.iii; 1.4x Mk.iii

awinphoto

  • 1D X
  • *******
  • Posts: 1991
    • View Profile
    • AW Photography
Re: Sensor Pixel Density
« Reply #2 on: September 21, 2011, 03:03:11 PM »
Thanks to you guys, I have a much better understanding of why a bigger sensor is better, and also why we don’t see super-zoom cameras with large sensors (because the lenses would have to be impractically large).

However, I have another sensor-related question or two:

I see a lot of complaining about the megapixel count being too high on some of the small-sensor cameras.  I can see why, as I have downloaded full resolution samples from say a 16 MP super-zoom, and the image is quite fuzzy at full resolution.  But the same picture looks quite good at my monitor’s resolution (1920 x 1080) and even zooming in a few levels.

Question 1:  If I were to set the resolution to a lower setting in the camera, would that solve the issue?  In other words would setting a 16MP camera to 10MP output produce the same image quality of a camera with a 10MP sensor assuming all other variables were the same? (yes, this is hypothetical since there probably are not two matching cameras with only the sensor pixel density being different)

Question 2: If the answer above is "no", would it be better or worse to reduce the image size in software on the computer?  (I know jpeg is destructive compression, so assuming you are not working with RAW images, I would think reducing the size on the computer would be worse since it would be going through Jpeg compression twice)

Thanks in advance for your insight.

Given that the 10MP would use the entire sensor vs a 15mp (50D) or 18mp (60D/7D) shot at a medium setting, You wouldn't get the entire goodness of the sensor... so the 10MP, in this example, would be optimum.  On a disclaimer, as far as raw sharpness goes, the 18mp 7d is not any softer than a 12mp 40D when using good lenses... The higher the MP, the more demanding it will be on the lenses, and quite frankly, it will expose the flaws and softness on cheaper/consumer grade lenses than the lower MP... The only part where the lower MP really becomes an advantage is diffraction on the lens, however between these two cameras you are looking at less than a stop difference.

Now if you were going to shoot 18MP and downsample to the same dimensions as the 10/12MP camera... the 18MP may be as good if not better because if you set photoshop to bicubic, it will come out quite nice and naturally sharpen... On tests vs the 7D and the nikon D300s, many reviews did the same thing (downsampled the 7D to match and then on the flip side up-sampled the nikon to match the 7D...)  It's not a real fair matchup but it is what it is...

What to take from this, dont fear the higher MP cameras if you can afford nice glass... Remember to invest more in your glass than your cameras and you should be fine. 
Canon 5d III, Canon 24-105L, Canon 17-40L, Canon 70-200 F4L, Canon 100L 2.8, 430EX 2's and a lot of bumps along the road to get to where I am.

Mega

  • Guest
Re: Sensor Pixel Density
« Reply #3 on: September 21, 2011, 06:29:32 PM »
K-amps, thanks, that does make sense about interpolated pixels appearing worse than native pixels.  However, in the case of the consumer grade point-and-shoot super-zooms that I am looking at, most do not support RAW which is the only way to get true native pixels.  So even a 16MP sensor saving to a 16MP file is going to get a little blurred from the noise reduction and jpeg compression, etc.

What to take from this, dont fear the higher MP cameras if you can afford nice glass... Remember to invest more in your glass than your cameras and you should be fine. 
I probably should have been clearer, but I was intentionally being vague as to not get into a discussion about specific cameras or brands.  But I am not currently in the market for anything with interchangeable lenses.  I am specifically looking for a super-zoom, and every single model I am considering right now has a 1/2.3-inch (tiny) sensor.

I am leaning towards the Sony DSC-HX100V mainly because of its superior movie recording (1080/60p).  But I see a lot of complaints that 16MP is just too many pixels for such a small sensor, to which my gut reaction is: “Well just change the file size if it’s too many.”  The Canon SX40 HS is pretty attractive except for its max movie mode being 1080/24p which I think will be noticeably worse for fast-moving subjects.  But its sensor is 12MP.  So I am just trying to figure out if 12MP is an actual advantage over 16MP (which seems counterintuitive) unless there is no way to get a 12MP or smaller image out of the Sony without losing some fidelity (which apparently may not be there to start with in either camera anyway).

In general, I have seen more than one “expert” suggest that 8MP is about as much as they should try to cram on a 1/2.3-inch sensor for a consumer grade camera.  But I am hard pressed to find a reason not to put more as long as there is no disadvantage (other than file size).

zhgart

  • Guest
Re: Sensor Pixel Density
« Reply #4 on: October 26, 2011, 06:42:44 AM »
sensor pixel density is different from sensor size and pixels number.
pixels number = density x chip area.

neuroanatomist

  • CR GEEK
  • ********
  • Posts: 14394
    • View Profile
Re: Sensor Pixel Density
« Reply #5 on: October 26, 2011, 09:25:42 AM »
Echoing others, setting a 16 MP camera to 10 MP would not deliver better resoluton, it's still just downsampling the image, and it would be better to do that in Photoshop.  But, would 4 MP be enough for you?  If the camera with a 16 MP sensor was designed to do so, a could bin 2x2 arrays of pixels and deliver a 4 MP image that would be superior to downsampling the 16 MP image to 4 MP. 

Ok, so 4 MP is too low, fine.  But extrapolate that to a 36 MP sensor (not that I'm naming a brand or a specific rumor or anything).  If such a camera were designed to do 2x2 binning, some very nice 9 MP images would result, especially in low light shooting situations, and those 9 MP images would be 'real' resolution without the demosaicing normally performed.  Now, if those 9 MP images were generated for customers who, hypothetically-speaking, were historically used to cameras with lower MP counts than competing brands, well...that might get interesting!
EOS 1D X, EOS M, and lots of lenses
______________________________
Flickr | TDP Profile/Gear List

canon rumors FORUM

Re: Sensor Pixel Density
« Reply #5 on: October 26, 2011, 09:25:42 AM »