Cascaded Face Sketch Synthesis under Various Illuminations.

Area of Science:

• Computer Vision
• Image Processing
• Artificial Intelligence

Background:

• Face sketch synthesis is crucial for digital entertainment but current methods struggle with varying lighting.
• Existing face sketch synthesis techniques are sensitive to illumination changes, leading to unsatisfactory results in real-world settings.

Purpose of the Study:

• To develop a robust face sketch synthesis framework that performs consistently under uncontrolled lighting conditions.
• To improve the quality and reliability of synthesized face sketches compared to existing methods.

Main Methods:

• Proposed a novel cascaded face sketch synthesis framework.
• Incorporated a multiple feature generator for artist-style sketch features and illumination-robust photo features.
• Utilized a cascaded low-rank representation to progressively reduce the difference between synthesized and real sketches.

Main Results:

• The framework generates realistic sketches comparable to current methods under controlled lighting.
• Demonstrated significantly improved performance on datasets with varied illuminations and web-scraped celebrity photos.
• The proposed method shows robustness to lighting variations, a key limitation in prior work.

Conclusions:

• The novel cascaded framework offers a robust solution for face sketch synthesis under diverse lighting conditions.
• This approach paves the way for advanced computer-aided optical systems in face sketch synthesis and imaging.
• The method addresses a critical need for reliable sketch synthesis in uncontrolled real-world environments.