CS vs. JPEG

This page is created for our paper "Image Compression Based on Compressive Sensing: End-to-End Comparison with JPEG"

Authors: Xin Yuan, and Raziel Haimi-Cohen

Abstract:

We present an end-to-end image compression system based on compressive sensing. The presented system integrates the conventional scheme of compressive sampling and reconstruction with quantization and entropy coding. The compression performance, in terms of decoded image quality versus data rate, is shown to be comparable with JPEG and significantly better at the low rate range. We study the parameters that influence the system performance, including (i) the choice of sensing matrix, (ii) the trade-off between quantization and compression ratio, and (iii) the reconstruction algorithms.We propose an effective method to jointly control the quantization step and compression ratio in order to achieve near optimal quality at any given bit rate. Furthermore, our proposed image compression system can be directly used in the compressive sensing camera, \eg the single pixel camera, to construct a compressive hardware sampling system.

--

We present more results on this page:

All these results can be reproduced by the code downloadable at CS vs. JPEG demo

The 200 images results from BSDS300 can be downloaded at https://drive.google.com/file/d/1sRO7mTmbQ6PaDTl7UfdlvGFPdFv0WgGG/view?usp=sharing

1) Image compression architecture comparison between proposed CSbIC (top) and JPEG (bottom).

2) SSIM vs. Compressed file size (in Bytes) for GAP-TV, D-AMP, NLR-CS

This figure is Figure 4 in the main paper

Performance diagrams, SSIM vs. compressed file size (in bytes), comparing JPEG (black solid curves), JPEG2000 (black dash curves) with CSbIC compression using different sensing matrices -- 2D-DCT (solid) and 2D-WHT (dash), and different reconstruction algorithms --- GAP-TV (blue), NLR-CS (red), and D-AMP (green). Top: 8 widely used image. bottom: 8 exemplar images from the BSDS 300 dataset.

3) PSNR vs. Compressed file size (in Bytes) for GAP-TV, D-AMP, NLR-CS. This includes all the 8 images while Figure 5 in the paper shows 2 image as examples

4) Example images with similar size to JPEG

5) Compare different sensing matrices using GAP-TV, D-AMP and NLR-CS

6) Compare reconstruction results with and without quantization