Sorry, you need to enable JavaScript to visit this website.

Metalenses are increasingly used in compact optical devices, such as AR/VR and quantum computing systems. This work introduces a complete design simulation pipeline for a metasurface system, from initially setting target phase profiles to the subsequent stage of choosing meta-atom material parameters and shape optimization, and finally characterizing optical performance. We demonstrate the functionality of our developed simulation platform on an RGB metalens system design.

This document contains the supplementary material accompanying the paper “Model-Aware Rate–Distortion Limits for Task–Oriented Source Coding,” published at the 2026 IEEE International Conference on Image Processing (ICIP 2026), Paper ID 875.

The supplementary material provides additional results, analyses and figures supporting the experiments and conclusions presented in the main paper.

Recipe generation from food images has practical value for intelligent cooking assistance, but traditional one-pass generation often overlooks user-specific safety requirements such as allergies, dietary restrictions, and preparation constraints. We propose \textbf{PCAR}, a \textbf{P}lanner-\textbf{C}ritic \textbf{A}gentic \textbf{R}emediation framework for trustworthy recipe generation.

Accurate detection of camouflaged crops, where objects share visual characteristics with their surrounding foliage, is critical for modern agricultural tasks. However, the high computational demands of state-of-the-art deep learning detectors often preclude their deployment on resource-constrained edge devices. This paper presents a systematic evaluation of three lightweight design strategies comprising backbone replacement, pruning, and knowledge distillation applied to SSD, YOLOv8l, and RT-DETR architectures.

Car sticker detection is a challenging small-object detection problem due to the tiny size of stickers, cluttered backgrounds, and limited annotated data. These challenges are further amplified in few-shot settings where only a small number of labeled instances are available for training. In this work, we propose Windshield-to-Sticker Detection (WiSDet), a region-guided detection pipeline that first localizes the windshield and then performs sticker detection within the cropped region to reduce background interference and improve effective sticker resolution.

I present an OCR-guided system for extreme license plate super-resolution from short degraded image sequences. Instead of selecting a single restored image, the method generates complementary restoration hypotheses using denoising, 4×resizing, deskewing, multi-frame alignment, fusion, low-rank SVD reconstruction, deconvolution, and outlier-disqualified averaging. These candidates are recognized by two fine-tuned FastPlateOCR CCT-XS-V2 models exported for ONNX Runtime inference, and the final plate string is obtained by reliability-weighted character voting.

Tube-sheet endoscopic inspection images pose a challenging defect localization problem because the valid inspection region occupies only part of the frame, while defects are typically small, localized, and visually subtle. This paper proposes an ROI-aware CLIP-based defect-localization framework for tube-sheet endoscopic inspection using ViP2-CLIP.

Unsupervised fine-grained image hashing aims to learn compact binary codes that preserve subtle visual differences among highly similar instances without manual annotations. However, most existing methods neglect collision resistance, leading to identical hash codes for slightly semantically different samples. In this paper, we propose Collision-Resistant Single-Pass Self-Supervised Semantic Hashing (CS3H), a collision-resistant framework that directly optimizes Hamming-space similarity via a single-pass normalized Hamming distance loss to produce well-separated binary representations.

Ancient Sanskrit manuscripts on ochre parchment represent a critical challenge for Document Image Analysis. Standard binarisation methods conflate chromatic staining with ink, leading to topological loss. We present MorphBG, a pipeline using per-channel morphological background estimation to preserve ink signals. Unlike recent deep learning methods, MorphBG is a classical, CPU-efficient approach. We demonstrate its scalability by processing over 1,200 manuscript pages. While MorphBG achieves a DRD of 2.38, zero-shot OCR engines yield 0% accuracy on this script.

This document contains the supplementary material to the
work “Manifold Optimization on the Magnitude Torus for
Fourier Phase Retrieval.”
In Sec. A, we present the proofs of the theorems and
corollaries appearing in the manuscript. In Sec. B, we provide
additional numerical simulations of the dualPGD algorithm
to demonstrate its applicability. Further remarks on our results
are given in Sec. C.