Abstract
The project aims at developing a new type of software products for lossless image compression
where the size of the compressed file is comparable with the one provided by standard lossless
compression products and, besides, the content of the compressed image remains visible and
completely intelligible. Since the content is visible, the decompression is necessary only when
the exact recovery of the original image is required. The idea of such compressors appeared in the
framework of the exploratory research project PN-III-P4-IDPCE-2016-0339, when we investigated the
improvement of lossless compression by using high capacity reversible data hiding (RDH) algorithms.
The basic principle of the proposed RDH-compressors is to reduce image size (by subsampling or
cropping) and to embed into the reduced size version, by RDH, the entire information needed to
exactly recover the original image. The concept, RDH-compressors, is new and is still unpublished.
The main objectives of the project are to design, develop and test prototypes for the
RDH modules and for the RDH-compressors both for graylevel and color images.
Our preliminary high capacity RDH schemes are already the state-of-the-art in data hiding
and can provide, for graylevel images, embedding bitrates of more than 3 bits per pixel.
Improvement of about 0.1 to 0.3 bpp are expected. Besides lossless compressors,
the high capacity RDH schemes can be used in complex metadata embedding applications
(e.g. embedding images into images). The preliminary RDH-compression rates are similar
with the ones of JPEG-LS. The expected prototypes will outperform the compression rates
of JPEG-LS and will be rather similar with the ones of PNG and JPEG 2000, with the advantage
of immediate access to image content, without any computation. The interest in RDH-compressors
is broad and covers any application where digital images are stored, retrieved, manipulated,
received or transmitted.
Objectives
Testing and validation of high capacity reversible data hiding (RDH) for graylevel images.
Design, testing and validation of lossless RDH-compressor prototype for graylevel images and its
corresponding RDH-decompressor.
Extension to color images of the high capacity RDH.
Extension of the lossless RDH-compressor prototype to color images, testing and validation.
Results
The RDH for graylevel images provides embedding bitrates greater than 3 bpp (3.27 bpp on graylevel Kodak set)
and, for color images, over 12 bpp (12.25 bpp on Kodak set). The lossless compressor not only provides compression,
but also the content of the image remains visible and completely intelligible. In terms of compression, the proposed
RDH-based lossless compressor outperforms JPEG-LS, TIFF, PNG and is outperformed by JPEG-2000.