Moderator: jsachs
The documentation of the API is here : https://www.libraw.org/docs/API-datastruct-eng.htmlLibRaw provides several options to manage how raw data is processed, which might help address the issue of highlight clipping. Specifically, you can explore the structure libraw_output_params_t, which includes parameters for managing post-processing. One key parameter is highlight, which allows you to control how highlights are handled during processing. Setting this parameter to a value like 0 (no clipping) or 1 (clip to white) might help you avoid clipping.
Additionally, you can look into the gamma settings or other raw decoding options in the libraw_pack_unpack_params_t structure to fine-tune the output.
float auto_bright_thr;
dcraw keys: none
Portion of clipped pixels when auto brightness increase is used. Default value is 0.01 (1%) for dcraw compatibility. Recommended value for modern low-noise multimegapixel cameras depends on shooting style. Values in 0.001-0.00003 range looks reasonable.
Im not sure of what your answer means.Could you enable user-control of highlight and gamma ? .... Been there done that.
darktable uses the RawSpeed library to read raw files, including linear DNGs. RawSpeed is an open-source library originally developed by Klaus Post and now maintained as part of the darktable project.
Here's how darktable potentially handles the aspects you mentioned regarding reading linear DNG raw files and managing highlight clipping:
1. Reading Linear DNG Raw Files with RawSpeed:
darktable relies on RawSpeed to decode the raw data from linear DNG files. RawSpeed supports a large number of raw file formats from various camera manufacturers, including the DNG format.
RawSpeed provides the unmodified raw data, possibly scaled to 16 bits or normalized to 0->1 floating-point data.
It also provides the CFA (Color Filter Array) layout for known cameras and automatically calculates the black level for cameras with this information.
2. Highlight Clipping Management (related to LibRaw):
Although darktable uses RawSpeed for reading raw files, it's helpful to understand how the concepts mentioned regarding LibRaw might translate in the context of darktable:
libraw_output_params_t and the highlight parameter: RawSpeed, like LibRaw, must handle highlight clipping during demosaicing and initial processing of the raw data. While darktable doesn't directly expose the libraw_output_params_t structure, it offers several modules and options to manage highlights:
The "highlight reconstruction" module offers different methods to recover details in overexposed areas. These methods include "clip highlights", "reconstruct in LCh", and "reconstruct color". The "clip highlights" option would correspond to behavior similar to setting the highlight parameter to 1 in LibRaw (clip to white). The other methods attempt to reconstruct color and brightness information in clipped areas using neighboring unclipped pixels.
The "filmic rgb" module (or the older "base curve" module) also has a significant impact on highlight management by applying a tone curve that can compress the dynamic range and reduce clipping. The "filmic rgb" module has a dedicated "reconstruct" section for managing highlight clipping.
The "exposure" module allows adjusting the overall exposure of the image, which can help bring highlights back into a non-clipped range.
The "overexposure indicator" module allows visualizing areas of the image that are clipped in different color channels, thus helping the user identify and correct highlight issues.
libraw_pack_unpack_params_t and the gamma settings and other raw decoding options: Again, RawSpeed handles the unpacking of raw data. darktable, at the user level, does not allow direct manipulation of low-level parameters like gamma at the time of raw file reading. However, darktable's processing pipeline includes modules such as:
"input color profile": This module manages the conversion from the camera's color space to a working space. There might be profile options that influence how colors are interpreted from the beginning of the pipeline.
"exposure": As mentioned earlier, it allows adjusting the brightness of the image linearly.
Other tone and color adjustment modules (such as "tone curve", "levels", "color equalizer") allow precise control over the final appearance of the image, including the management of highlights and shadows.
To find the specific code in darktable that interacts with RawSpeed to read linear DNGs and potentially influence highlight management, you should explore the darktable sources on GitHub, especially the files in the following directories (keeping in mind that RawSpeed is an external library):
src/imageio/: This directory contains the code for managing image input and output, including the integration with RawSpeed. Look for files related to reading raw formats and specifically DNG.
src/processing/: This directory contains the code for the various image processing modules, including "highlight reconstruction", "filmic rgb", and "exposure". Examine how these modules analyze and modify image data, especially concerning high-brightness pixels.
The RawSpeed sources themselves (in the darktable-org/rawspeed repository) might contain detailed information about how linear DNG data is interpreted and how initial clipping is handled.
In summary:
darktable uses RawSpeed to read linear DNG raw files. While it doesn't directly expose the libraw_output_params_t and libraw_pack_unpack_params_t structures, darktable offers a variety of modules and options in its processing pipeline to effectively manage highlight clipping and adjust the tone and colors of raw images, including those from linear DNG files. Exploring the source code of darktable and RawSpeed on GitHub will give you a deeper understanding of the specific mechanisms involved.