基於巨觀邊緣感知與對比圖分析的高動態範圍成像

No Thumbnail Available

Date

2025

Journal Title

Journal ISSN

Volume Title

Publisher

Abstract

本研究主要探討高動態範圍影像處理中,在極端光照條件下的影像主體重建與色調映射問題。當畫面遭遇強烈背光、低光源或大面積背景干擾時,傳統的影像處理方法經常無法有效地凸顯主體,且在壓縮動態範圍過程中容易造成細節喪失與色彩失真。本研究針對上述問題,提出一套考慮巨集邊緣資訊的影像處理方法,結合全域與區域對比度評估,透過調整適合的色調映射曲線,使暗部細節清晰且避免亮部過曝。此外,研究中亦針對色彩還原問題,於色彩轉換過程中引入色域映射模型的補償機制,有效避免傳統方法常見的色相偏移與飽和失真現象。本論文透過實驗驗證所提出方法的有效性,並經由業界常見的客觀指標評估其在亮度、色彩準確性與視覺對比度上的改善效果,期望能提供未來影像訊號處理系統設計的重要參考。
This study addresses subject reconstruction and tone mapping in HDR images under extreme lighting—strong backlight, low illumination, or large-area distractions. We propose a framework that leverages macro-edge information and combines global and local contrast evaluation to adapt tone-mapping curves, preserving shadow detail without over-exposing highlights. To improve color fidelity, an OKLab-based compensation is applied during color conversion, reducing hue shifts and saturation errors. Experimental results and VCX metrics confirm notable gains in luminance reconstruction, color accuracy, and visual contrast, offering a practical reference for future ISP system design.

Description

Keywords

高動態範圍, 色調重現, 色域映射, 對比增強, HDR, Tone Reproduction, Tone Mapping, Contrast Enhancement

Citation

Collections

Endorsement

Review

Supplemented By

Referenced By