Abstract

An image compression architecture and method implement a custom Discrete Wavelet Transform (DWT) for Gaussian-pyramid-based Image Signal Processing (ISP) pipelines. To reduce memory footprint and dynamic random-access memory (DRAM) bandwidth demands while mitigating the ringing artifacts of standard wavelets, a custom biorthogonal filter bank utilizes a low-pass analysis filter that directly implements a Gaussian smoothing response. Low-frequency sub-bands at each decomposition level directly correspond to matching Gaussian pyramid levels, enabling direct downstream ISP processing with little or no algorithmic redesign. The critically sampled transform preserves rectangular image dimensions in memory and executes via a multiplierless lifting scheme using fixed-point additions and binary shifts.

Creative Commons License

Creative Commons License
This work is licensed under a Creative Commons Attribution 4.0 License.

Share

COinS