Leica Showcase Leica M Monochrom

I get the feeling that they are swamped. But- How many camera manufacturers would offer a free service like this? It required a Lawsuit for Nikon to admit to the problem and replace the D600 shutters that sprayed oil on the sensors.


Bakersfield, CA
Real Name
I get the feeling that they are swamped. But- How many camera manufacturers would offer a free service like this? It required a Lawsuit for Nikon to admit to the problem and replace the D600 shutters that sprayed oil on the sensors.

My sensor is fine, just needs the rangefinder adjusted.


Hall of Famer
Real Name
Monochrome 246 with zeiss 35mmf2 SOOC
75 year difference between youngest and oldest.

Nikkor 13.5cm F3.5, Wide-Open, ISO 10,000. No noise reduction, used my custom software to add a Gamma curve to the DNG file.

First time I've tried the 135 focal length on the Leica for skating. The ISO 10,000 on the M Monochrom makes this possible.
View attachment 192465

I noticed an intermittent dark line when shooting at High ISO- mostly occurring in dark areas. Some investigation, and custom code used to apply the Gamma curve- the pixel in Column 3203 row 3105 is not passing the full charge as the image is being read-out. The pixel is not dead, but is "non-Uniform". I'm going to add an algorithm to add an offset to the values in column 3203, pixels 1:3105 which show the subtraction of a DC offset. I'm going to compare the average values for the affected pixels with those on each side of it and add the offset to the values. I plotted columns 3202, 3203, and 3204. All of the detail is preserved in the bad column, it's just offset from where it should be. I suspect if the camera was sent to Leica they would map out the bad column and the average of the two neighboring columns would be used to fill in the image. Since I process the DNG files with my own software anyway, I'll add the non-uniformity correction to the code.
Last edited:
The non-uniformity is not just a DC offset, but also has a non-linear component to it, probably relating to "charge transfer efficiency" being different on the bad pixel. I ended up computing the local difference from the pixel from the moving average of five pixels of neighbors and adding it to the corrupt value. Not perfect, but better than averaging the values. For the time being, my code leaves the original DNG file intact, applies the change to the gamma DNG files that it generates.
Last edited:

Here's the output from this mornings work- used a moving average of 5 values for columns on each side of the bad one, and for the bad column. Added some code to compute thresholds used to decide if a pixel gets changed or not- the error is dependent on ISO used, and on the intensity values of the image. The difference of the moving average is added to the pixel, a cheap way of computing a local DC offset. Nina wanted to know what I was doing, told her it's just like the non-uniformity corrections we worked on together for a digital sensor in 1983.

Just to add- did a 3:1 zoom of pixels, comparing the original with dark line to the "Local DC Level adjusted image". My eye could not pick up the mending in the image, it's that close. There is structure in the dark-line, just a value subtracted out by the bad pixel. Add that value back to the affected pixels: blends in, better than replacing the line with the average of the pixels on each side of it.

Of course my software runs under DOS and is written in FORTRAN-77, as it is modified from code written in the early 1980s.

I found a recent article on restoring pixel values from columns that have been "dragged down" as happened on my camera.

The article does a good job of listing types of "bad Pixels", mine was the first type:
• Linear pixels with false bias: a category where the pixel output behaves linearly with illumination intensity or exposure time, but due to either a low or high bias compared to its neighbours, values displayed by this pixel differs considerably from the ones around it, given an homogeneous illumination;
• Nonlinear pixels: a subset of defect pixels, where the output signal displayed on an image does not relate to the illumination intensity or exposure time in a linear fashion;
• Dead pixels: are those who have very low sensitivity to illumination intensity variations, consistently presenting low grey values;
• Hot pixels: this category is highly sensible to illumination intensity variations, presenting persistent high grey values.


The article describes an algorithm to properly correct the defective values, most likely required for radiometric calibrated images.

My simple algorithm "just works" because the image cannot change so rapidly from one column to another that a moving average cannot be used to compute the error reasonably well.
Last edited:

Latest threads

Top Bottom