Root Delta Image (RDI): A Low-Complexity Lossy Still-Image Format with Scanline Root-Delta Prediction

Modern image codecs achieve impressive compression ratios, but their bitstreams and decoding pipelines have grown complex enough that implementing and hardening a decoder is often more challenging than the image-processing task that motivated the decoder in the first place.This work presents Root Delta Image (RDI) , a compact, intentionally low-complexity lossy raster format designed around (i) a reversible integer color decorrelation transform (YCoCg), (ii) per-scanline prediction with a constrained 4-bit root-delta quantizer, (iii) optional 2\((\times)\)2 chroma subsampling, and (iv) a file layout consisting of a fixed-size header plus exactly one \texttt{zlib}/DEFLATE-compressed payload.RDI trades compression efficiency for analyzability: the format has explicit size bounds and deterministic decode procedures that are expressible with bounded loops, integer arithmetic, and a single call to a commodity DEFLATE implementation.We summarize the format specification, describe a reference implementation, and evaluate two practical nibble-packed modes on natural images.Across the Kodak image suite and two TestImages.org subsets, RDI achieves perceptual similarity comparable to baseline JPEG at common quality settings while typically requiring higher bitrates; at the same time, it yields substantial size reductions compared to lossless PNG.We position RDI among predictive and transform-based codecs and highlight properties that are particularly relevant to embedded and security-sensitive deployments: compact decoding logic, predictable memory use, and a validation-first bitstream structure.