Prompt Description

Please generate a diagram of the core module architecture that complies with SCI journal illustration specifications. The requirements are as follows: 1. Overall Layout and Specifications: - Layout: Horizontal serial flow (adapted to a paper's single-column width of 8.5cm / double-column width of 17cm), with the structure: "Degraded Input Image" → Core Module 1 → Core Module 2 → Core Module 3 → "Restored Output Image"; - Color Scheme: Use low-saturation professional colors for the module backgrounds (Module 1: light blue, Module 2: light green, Module 3: light gray), deepen the sub-components by 20% within the same color family, and use a 0.5pt thin black line for the borders. - Font: Use Times New Roman for English (10pt bold for module titles, 8pt for sub-components/descriptions), and Songti for Chinese (same font size levels); label feature variables (e.g., F) in italics. - Resolution: ≥300dpi, vector format (supports scaling without blurring). 2. Module Details (including internal sub-components): - Module 1: Title (bold) "Core Module 1: Degradation-Aware Dynamic Routing Module" Embedded sub-component boxes (darker light blue): ① Degradation Type Recognition Unit: Function description "Extract degradation representations (e.g., blur/noise) from input image"; ② Adaptive Decision Mechanism: Function description "Classify degradation categories via learnable thresholds"; ③ Dynamic Routing Strategy Library: Function description "Match corresponding restoration strategies for multi-type degradation"; Data flow annotation: Input feature "*F*degraded" → Module 1 → Output feature "*F*1". - Module 2: Title (bold) "Core Module 2: Multi-Modal Feature Injection Module" Embedded sub-component boxes (darker light green): ① CLIP/LLM Semantic Prior Extractor: Function description "Retrieve textural semantic priors from pre-trained models"; ② Cross-Modal Feature Alignment Unit: