IQ-LUT: interpolated and quantized LUT for efficient image super-resolution

TL;DR

IQ-LUT introduces an efficient LUT-based method for image super-resolution that reduces storage requirements by up to 50 times while improving visual quality through integrated interpolation, quantization, residual learning, and knowledge distillation.

Contribution

The paper presents IQ-LUT, a novel approach combining interpolation, quantization, residual learning, and knowledge distillation to significantly reduce LUT size and enhance super-resolution quality.

Findings

LUT size reduced by up to 50x compared to ECNN.

Super-resolution quality surpasses previous methods.

Storage costs are substantially decreased while maintaining high quality.

Abstract

Lookup table (LUT) methods demonstrate considerable potential in accelerating image super-resolution inference. However, pursuing higher image quality through larger receptive fields and bit-depth triggers exponential growth in the LUT's index space, creating a storage bottleneck that limits deployment on resource-constrained devices. We introduce IQ-LUT, which achieves a reduction in LUT size while simultaneously enhancing super-resolution quality. First, we integrate interpolation and quantization into the single-input, multiple-output ECNN, which dramatically reduces the index space and thereby the overall LUT size. Second, the integration of residual learning mitigates the dependence on LUT bit-depth, which facilitates training stability and prioritizes the reconstruction of fine-grained details for superior visual quality. Finally, guided by knowledge distillation, our non-uniform quantization process optimizes the quantization levels, thereby reducing storage while also compensating for quantization loss. Extensive benchmarking demonstrates our approach substantially reduces storage costs (by up to 50x compared to ECNN) while achieving superior super-resolution quality.