notesum.ai

Published at December 10

Deep Joint Unrolling for Deblurring and Low-Light Image Enhancement (JUDE).pdf

cs.CV

Released Date: December 10, 2024

Authors: Tu Vo, Chan Y. Park

Arxiv: http://arxiv.org/pdf/2412.07527v1

Refer to caption

Category	Methods	Param. (M)	PSNR $\uparrow$	SSIM $\uparrow$	LPIPS $\downarrow$
LLE →Deblur	FourLLIE →FFTFormer	16.72	18.433	0.705	0.305
	LLFormer →FFTFormer	39.12	20.290	0.792	0.212
	RetinexFormer →FFTFormer	18.21	16.452	0.702	0.324
Deblur →LLE	MIMO →RetinexFormer	17.71	17.024	0.770	0.271
Deblur →LLE	FFTFormer →RetinexFormer	18.21	16.712	0.728	0.325
LLE & Deblur	FFTFormer	16.6	19.889	0.858	0.139
	RetinexFormer	1.61	25.505	0.862	0.240
	LEDNet [58]	7.5	25.740	0.850	0.224
	FELI [12]	0.6	26.728	0.914	0.132
	JUDE (Proposed)	35.2	26.884	0.932	0.127