
In raw imaging workflows, metadata is as critical as pixel data. Raw files do not merely store sensor measurements; they also encode the information required to interpret those measurements as color, brightness, and dynamic range. When computational methods modify raw-domain image data, this interpretive metadata is often lost, invalidated, or omitted, producing substantial rendering inconsistencies in downstream software. PARSEK (the Probabilistic Alignment Raw Stitcher Experiment from Kentucky) exposes this problem clearly: although its super-resolution output preserves useful raw-domain image content, results saved without appropriate metadata can exhibit severe color and tonal shifts when opened in conventional raw processors. This paper presents KYDNG, a DNG repackaging pipeline designed to preserve perceptual consistency by embedding reconstructed raw image data together with metadata derived from the source capture. The implementation writes raw-specific structural tags, including CFA pattern and repeat dimensions, and propagates key camera-dependent rendering metadata, including ColorMatrix1, AsShotNeutral, and black and white levels, while also embedding a JPEG preview. The resulting files are intended to be interpreted by standard raw development tools using the same camera-consistent rendering assumptions as the original capture.

For cellphone and camera manufacturers, App developers and test labs it is very helpful to know how cameras are used. Under which lighting conditions are images captured? What is captured? Was the flash used? Many more things can be analyzed by looking at the image content and the metadata. The last analysis dates 9 years back to a time when smart phones with integrated cameras started to take off. Since then a lot of the usage conditions of image capture devices have changed. People don't use the flashes anymore, they take a lot more pictures of food than before and a lot of the images are selfies that did not exist. Knowing how a camera is used can help prioritize certain developments, allows to optimize algorithms for image processing and can help archiving images in better ways.