EUVIMEDEUVIMED entdecken
Open Access · CC BY 4.0

EQISP: efficient quantized image signal processing with multi-scale pyramid fusion for resource constrained embodied perception

Tongxin Yang, Ling Guo, Jie Li, Qi Qin, Bowen Liu, Qian Zhang

Frontiers in Neurorobotics · 08.06.2026 · 20

DOI: 10.3389/fnbot.2026.1866481 ↗

Kurz gefasst

Fachartikel in der Zeitschrift Frontiers in Neurorobotics, online erschienen am 08.06.2026. Verfasst von Tongxin Yang und 5 weiteren Autor:innen. Beteiligte Einrichtung: School of Naval Architecture, Ocean and Civil Engineering, Shanghai Jiao Tong University, Shanghai, China. Die Arbeit nennt 48 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0).

AbstractSchlagwörterBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Introduction

Resource-constrained environmental perception requires autonomous robots and embodied intelligent systems to process visual signals efficiently while preserving image fidelity in complex real-world environments. However, converting high dynamic range RAW sensor data into perceptually faithful RGB images remains computationally expensive, thereby limiting the deployment of neural image signal processors on edge platforms with restricted memory, energy, and computational budgets.

Methods

Consequently, this study proposes the enhanced quantized image signal processor (EQISP), comprising the quantized convolutional neural network (QCNN) and the unified pyramid fusion algorithm (UPFA). QCNN employs dynamic fixed-point hybrid quantization, which adjusts parameter ranges according to the linear relationship between threshold standard deviation and fractional length, thereby significantly reducing the computational load. Meanwhile, UPFA utilizes Gaussian pyramids to capture global illumination and Laplacian pyramids to preserve fine details, enabling multi-scale, multi-exposure fusion and iterative reconstruction to mitigate detail loss induced by quantization.

Results

Comprehensive comparative experiments demonstrated that EQISP achieved a PSNR of 22.90 dB, an SSIM of 0.9278, and 164.843 GFLOPs. Compared with the PyNET baseline, EQISP improved the PSNR by 1.71 dB while reducing the computational cost by a factor of 4.24. Furthermore, deployment experiments on an NVIDIA Jetson TX2 development board showed that EQISP achieved a model size of 57 MB, an inference latency of 189 ms, an inference speed of 6.1 FPS, and a peak memory usage of 2.2 GB.

Discussion

These results provide practical evidence that EQISP can serve as an efficient and scalable visual front end for resource-constrained embodied perception systems.

Schlagwörter

Stichwörter der Autor:innen

AI-ISPedge visionembodied intelligent systemsembodied perceptionpyramid fusionquantization

Autor:innen und Institutionen

  • School of Naval Architecture, Ocean and Civil Engineering, Shanghai Jiao Tong University, Shanghai, China.
  • The School of Computer Science and Engineering, Chongqing University of Science and Technology, Chongqing, China.
  • Department of Military Logistics, Engineering University of the PLA Joint Logistic Support Force, Chongqing, China.
  • School of Computer Science and Internet of Things, Chongqing Institute of Engineering, Chongqing, China.

Zitieren

APA

Yang, T., Guo, L., Li, J., Qin, Q., Liu, B., & Zhang, Q. (2026). EQISP: efficient quantized image signal processing with multi-scale pyramid fusion for resource constrained embodied perception. Frontiers in Neurorobotics, 20. https://doi.org/10.3389/fnbot.2026.1866481

Vancouver

Yang T, Guo L, Li J, Qin Q, Liu B, Zhang Q. EQISP: efficient quantized image signal processing with multi-scale pyramid fusion for resource constrained embodied perception. Frontiers in Neurorobotics. 2026;20. doi:10.3389/fnbot.2026.1866481

BibTeX
@article{Yang2026,
  author = {Yang, Tongxin and Guo, Ling and Li, Jie and Qin, Qi and Liu, Bowen and Zhang, Qian},
  title = {EQISP: efficient quantized image signal processing with multi-scale pyramid fusion for resource constrained embodied perception},
  journal = {Frontiers in Neurorobotics},
  year = {2026},
  volume = {20},
  doi = {10.3389/fnbot.2026.1866481}
}
Kostenlos starten →