willie_901
Veteran
To address the original poster's comments:
1. There is no substitute for high quality data. Period.
In my view the 4/3 sensor size dictates that large angle-of-view lens production is prohibitively expensive. The only practical way to achieve competitive image quality is to fix the problems with post processing models.
Leica faced a different form of this problem with the M8, Leica's only option was to filter the analog light (the data is not modified!). This also true for the mis-named noise reduction software. The noise is the uncertainty in the data. It can not be reduced, it can only be filtered by averaging it with data that is assumed to have less uncertainty. This is a practical compromise, but the pixels with less noise are degraded by averaging them with pixels with more noise. However you will sell more software by only telling half the story calling it noise reduction, rather than the whole story by calling it data averaging.
2. Information theory tells us that original data should never be modified. This is true for everything from radio astronomy, to MRI images, to RAW image files. The less the camera vendor does to the RAW data the better. This has nothing to do with aesthetics, or marketing or manufacturing costs. It has to do with the fundamental nature of information.
Common sense tell us that a jpeg image out of the camera must have less information content than the RAW image used to create the jpeg. Many people decide the convenience of the jpeg is more important than the lost information content. This is not to say important photography can not be done using jpegs from the camera. But the same principle applies to in camera manipulated RAW data.
Of course, after data collection, one may do whatever one likes to model the data. In the case of the G1, the OEM lens imperfections are modeled and the data is modified to minimize those imperfections. How well the model maps onto reality determines the quality of the result. Here reality refers to the real, but unknown, information content of the light reflected from the subject into the lens. If the RAW file is modified to model the mean OEM lens distortions and your lens (by the luck of the draw) happens to have five time less distortion than the mean, then your lens' performance will be degraded by the model.
If an improved model for the OEM lens distortion becomes available, then all the previous RAW files can never use the new model, as they were modified.
The modifications to the RAW data collected with OEM lenses are proprietary. This means the only way we can assess how well the model estimates the real, but unknown light is subjective. Apparently the data from non-OEM lenses are modified less than than data from OEM lenses. Obviously some people don't like the model for non-OEM RAW data while others do.
I can understand how a company could modify the RAW data to minimize the characteristics of lenses that they never dreamed would be used with their in-camera firmware. Perhaps they feel that third-party lens manufacturers like Sigma and Tokina will not be able to compete if the RAW files are modified.
1. There is no substitute for high quality data. Period.
In my view the 4/3 sensor size dictates that large angle-of-view lens production is prohibitively expensive. The only practical way to achieve competitive image quality is to fix the problems with post processing models.
Leica faced a different form of this problem with the M8, Leica's only option was to filter the analog light (the data is not modified!). This also true for the mis-named noise reduction software. The noise is the uncertainty in the data. It can not be reduced, it can only be filtered by averaging it with data that is assumed to have less uncertainty. This is a practical compromise, but the pixels with less noise are degraded by averaging them with pixels with more noise. However you will sell more software by only telling half the story calling it noise reduction, rather than the whole story by calling it data averaging.
2. Information theory tells us that original data should never be modified. This is true for everything from radio astronomy, to MRI images, to RAW image files. The less the camera vendor does to the RAW data the better. This has nothing to do with aesthetics, or marketing or manufacturing costs. It has to do with the fundamental nature of information.
Common sense tell us that a jpeg image out of the camera must have less information content than the RAW image used to create the jpeg. Many people decide the convenience of the jpeg is more important than the lost information content. This is not to say important photography can not be done using jpegs from the camera. But the same principle applies to in camera manipulated RAW data.
Of course, after data collection, one may do whatever one likes to model the data. In the case of the G1, the OEM lens imperfections are modeled and the data is modified to minimize those imperfections. How well the model maps onto reality determines the quality of the result. Here reality refers to the real, but unknown, information content of the light reflected from the subject into the lens. If the RAW file is modified to model the mean OEM lens distortions and your lens (by the luck of the draw) happens to have five time less distortion than the mean, then your lens' performance will be degraded by the model.
If an improved model for the OEM lens distortion becomes available, then all the previous RAW files can never use the new model, as they were modified.
The modifications to the RAW data collected with OEM lenses are proprietary. This means the only way we can assess how well the model estimates the real, but unknown light is subjective. Apparently the data from non-OEM lenses are modified less than than data from OEM lenses. Obviously some people don't like the model for non-OEM RAW data while others do.
I can understand how a company could modify the RAW data to minimize the characteristics of lenses that they never dreamed would be used with their in-camera firmware. Perhaps they feel that third-party lens manufacturers like Sigma and Tokina will not be able to compete if the RAW files are modified.