
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.