Fractal image compression achieves high compression ratios but suffers from prohibitively long encoding times and limited reconstruction quality. To address these limitations, we propose fast fractal image compression using non-uniform partition (FFICNUP), a hybrid algorithm that adaptively partitions range blocks (R-blocks) and domain blocks (D-blocks) based on local texture and edge content. Smaller R-blocks are employed in texture-rich regions or edge-dense areas to preserve fine details, while larger R-blocks are adopted in smooth regions to accelerate encoding. By integrating a Task-Serial Workflow with Data-Parallel Vectorization and adaptive block partitioning, FFICNUP substantially accelerates both encoding and decoding processes while enhancing reconstruction fidelity and compression ratios. Experimental results demonstrate that the proposed FFICNUP method significantly outperforms conventional fractal image compression (FIC) approaches. By leveraging vectorized parallelization, the proposed FFICNUP achieves state-of-the-art (SOTA) decoding speed with a 14× acceleration, reduces encoding latency by three orders of magnitude, improves the Peak Signal-to-Noise Ratio (PSNR) by up to 5.19 dB, and attains a compression ratio 2.21 times higher than that of conventional FIC. Validated across both CPU and GPU platforms, FFICNUP dynamically balances encoding speed, reconstruction quality, compression ratio, and latency across varying image sizes, demonstrating its suitability for practical engineering applications.
Li et al. (Wed,) studied this question.