supplierdeeply

Bad pixels

  • 28 Replies
  • 14129 Views
*

Offline ursamajor

  • **
  • 61
  • SX150IS, SX510HS
    • Per aspera ad astra...
Bad pixels
« on: 27 / December / 2014, 06:38:38 »
Advertisements
I was horrified to discover that CHDK counted ab. 300000 bad pixels on my sensor.  :'(
Is there a way to phisically restart/fix bad (stuck/hot/dead) pixels in CMOS sensors? I remember that for my Canon 550D dSLR it was a method, by simply cleaning the sensor.
SX150IS (retired) :D, SX510HS

*

Offline srsa_4c

  • ******
  • 3686
Re: Bad pixels
« Reply #1 on: 27 / December / 2014, 09:02:03 »
I was horrified to discover that CHDK counted ab. 300000 bad pixels on my sensor.  :'(
Is there a way to phisically restart/fix bad (stuck/hot/dead) pixels in CMOS sensors? I remember that for my Canon 550D dSLR it was a method, by simply cleaning the sensor.
The cameras have the necessary firmware routines, but we don't know how to use them - only Canon services have the necessary PC-side software.

On the other hand, the bad pixel count CHDK reports is the number of pixels already considered bad by the firmware. You only need to worry if you see bad pixels on camera JPEGs.

*

Offline ursamajor

  • **
  • 61
  • SX150IS, SX510HS
    • Per aspera ad astra...
Re: Bad pixels
« Reply #2 on: 27 / December / 2014, 09:27:16 »
Indeed, no bad pixels in jpegs. Only in DNG 1.3 with some RAW edit softwares as RAW Therapee, ufraw etc. No bad pixels with DNG 1.1, so I switched on v. 1.1. Anyway, Adobe lightroom made the corrections with both DNG versions 1.1 and 1.3.
SX150IS (retired) :D, SX510HS

*

Offline reyalp

  • ******
  • 11487
Re: Bad pixels
« Reply #3 on: 27 / December / 2014, 15:10:10 »
Indeed, no bad pixels in jpegs. Only in DNG 1.3 with some RAW edit softwares as RAW Therapee, ufraw etc. No bad pixels with DNG 1.1, so I switched on v. 1.1. Anyway, Adobe lightroom made the corrections with both DNG versions 1.1 and 1.3.
Recent versions of raw therapee have a "dead pixel filter" option in the may do the right thing with DNG 1.3

You can also fix them with chdkptp but this is not very user friendly https://www.assembla.com/spaces/chdkptp/wiki/DNG_Processing

If CHDK actually counted 300,000 the DNG 1.1 badpixel process will take a lot of memory, which might causes crashes. If that happens, you should use 1.3. Even if that is a typo and the actual number is 30,000, that's still quite a lot.
Don't forget what the H stands for.


*

Offline ursamajor

  • **
  • 61
  • SX150IS, SX510HS
    • Per aspera ad astra...
Re: Bad pixels
« Reply #4 on: 28 / December / 2014, 04:56:09 »
I asked if there is a method to phisycally remove/restart bad pixels. I know how to remove them by software.
SX150IS (retired) :D, SX510HS

*

Offline philmoz

  • *****
  • 3070
    • Photos
Re: Bad pixels
« Reply #5 on: 28 / December / 2014, 05:12:24 »
I asked if there is a method to phisycally remove/restart bad pixels. I know how to remove them by software.

No, bad pixels can't be physically fixed short of replacing the sensor.

The sensor cleaning trick on DSLR cameras does not fix bad pixels; it just adds new ones to the mapping table so they are cleaned up in future images.

Phil.
CHDK ports:
  sx30is (1.00c, 1.00h, 1.00l, 1.00n & 1.00p)
  g12 (1.00c, 1.00e, 1.00f & 1.00g)
  sx130is (1.01d & 1.01f)
  ixus310hs (1.00a & 1.01a)
  sx40hs (1.00d, 1.00g & 1.00i)
  g1x (1.00e, 1.00f & 1.00g)

*

Offline ursamajor

  • **
  • 61
  • SX150IS, SX510HS
    • Per aspera ad astra...
Re: Bad pixels
« Reply #6 on: 28 / December / 2014, 11:37:35 »
AFAIK, some stuck pixels can be unlocked by long exposing the sensor. I used myself this method (exposing 3 times 5 minutes) and now CHDK counted "only" 130008 bad pixels from 300000. Hmmm! There is a decrease from 2.5% bad pixels to 1.08%. I wonder what's the maximum limit of bad pixels accepted by the manufacturer?
SX150IS (retired) :D, SX510HS

Re: Bad pixels
« Reply #7 on: 28 / December / 2014, 11:54:23 »
The number of bad pixels is temperature dependent.  You will get more bad ones as the sensor gets hotter.


*

Offline philmoz

  • *****
  • 3070
    • Photos
Re: Bad pixels
« Reply #8 on: 28 / December / 2014, 16:27:38 »
AFAIK, some stuck pixels can be unlocked by long exposing the sensor. I used myself this method (exposing 3 times 5 minutes) and now CHDK counted "only" 130008 bad pixels from 300000. Hmmm! There is a decrease from 2.5% bad pixels to 1.08%. I wonder what's the maximum limit of bad pixels accepted by the manufacturer?

CHDK only deals with dead pixels that always return a value of 0.
While stuck pixels (that always return a non-zero value) are technically 'bad' we don't cater for them.

A long exposure is more likely to show stuck pixels; but very unlikely to magically fix them.

Did you use a lower ISO the second time you ran the bad pixel detection in CHDK?
More dead pixels will often show using a high ISO.

If you want to use the built in bad pixel interpolation in CHDK (DNG 1.1) with the bad pixel list detected at low ISO, then when you shoot at high ISO you may end up with pixels not being corrected. It's better to use the bad pixel list detected at high ISO - the cleanup code will ignore pixels in image that are not 0 even though they are in the list.

Phil.
CHDK ports:
  sx30is (1.00c, 1.00h, 1.00l, 1.00n & 1.00p)
  g12 (1.00c, 1.00e, 1.00f & 1.00g)
  sx130is (1.01d & 1.01f)
  ixus310hs (1.00a & 1.01a)
  sx40hs (1.00d, 1.00g & 1.00i)
  g1x (1.00e, 1.00f & 1.00g)

*

Offline srsa_4c

  • ******
  • 3686
Re: Bad pixels
« Reply #9 on: 28 / December / 2014, 16:59:26 »
AFAIK, some stuck pixels can be unlocked by long exposing the sensor. I used myself this method (exposing 3 times 5 minutes) and now CHDK counted "only" 130008 bad pixels from 300000. Hmmm! There is a decrease from 2.5% bad pixels to 1.08%. I wonder what's the maximum limit of bad pixels accepted by the manufacturer?

CHDK only deals with dead pixels that always return a value of 0.
I think those "dead pixels" are deliberately made black by the firmware (depending on several factors - exposure time, temperature, ...). They are marked so to notify the JPEG engine that they need to be corrected (by interpolation). And I also believe that the original pixel values are not necessarily wrong. Attached is a (magnified) corner of a wide angle raw image taken by a CMOS sensor camera. The dark border is left as-is by the image pre-processing pipeline, marked bad pixels only appear in the inner area that is also corrected for vignetting. (In case someone is wondering - the large coloured spots are the result of demosaicing.)

 

Related Topics