High-fidelity Image Compression for High-throughput and Energy-efficient Cameras

Abstract

<p>In this thesis, we propose a new encoder-friendly image compression strategy for high-throughput cameras and other scenarios of resource-constrained encoders. The encoder performs L_∞-constrained predictive coding (DPCM coupled with uniform scalar quantizer), while the decoder solves an inverse problem of L₂ restoration of L_∞-coded images. Although designed for minimum encoder complexity (lower than distributed source coding and compressive sensing), the new codec outperforms state-of-the-art encoder-centric image codecs such as JPEG 2000 in PSNR for bit rates higher than 1.2 bpp, while maintaining much tighter L_∞ error bounds as well. This is achieved by exploiting the tight error bound on each pixel provided by the L_∞-constrained encoder and by locally adaptive image modeling.</p>