光学精密工程2025，Vol.33Issue(18)：2980-2995,16.DOI:10.37188/OPE.20253318.2980
高效Mamba驱动的端到端光场图像压缩
Efficient mamba-driven end-to-end light field image compression
摘要
Abstract
Light field images capture both spatial and angular information of light rays,providing richer vi-sual information than traditional 2D images.However,their high-dimensional nature poses challenges for existing compression methods in terms of global feature utilization,long-range correlation exploration,and computational complexity,limiting the improvements of compression performance and efficiency.To ad-dress these issues,this paper proposed an efficient Mamba-driven end-to-end light field image compression method.Firstly,2D slices containing spatial and epipolar plane information were extracted from the 4D light field image,and Mamba was employed to fully capture their global contextual information.Second-ly,to scan the light field image in multiple directions while avoiding a significant increase in computational complexity,a channel-efficient 2D selective scanning strategy was introduced to extract light field features accurately and efficiently.Finally,on the decoding end,a residual reconstruction module was designed to enhance the reconstructed image quality while reducing the number of parameters and decreasing the encod-ing and decoding time.The experimental results show that compared with the existing representative method SADN,the proposed method achieves an average bitrate reduction of 7.4%and a PSNR improve-ment of 0.37 dB on light field images with a 7×7 angular resolution,while also demonstrating superior subjective visual quality.In terms of encoding and decoding time,the proposed method has achieved a sig-nificant improvement of 10 to 20 times.Furthermore,compared to the state-of-the-art method LFIC-DRASC,the proposed method achieves an average bitrate reduction of 19.5%and a PSNR improvement of 0.58 dB on light field images with a 13×13 angular resolution.关键词
光场/图像压缩/端到端/MambaKey words
light field/image compression/end-to-end/Mamba分类
计算机与自动化引用本文复制引用
封哲宇,蒋志迪,万立飞,徐海勇,蒋刚毅..高效Mamba驱动的端到端光场图像压缩[J].光学精密工程,2025,33(18):2980-2995,16.基金项目
国家自然科学基金项目(No.62271276,No.62171243) （No.62271276,No.62171243）
