170 likes | 292 Vues
This study introduces an exemplar-based method aimed at automating the visual editing and retouching processes for fine art reproduction. Fine art reproduction is crucial, allowing rare and delicate artworks to be captured and displayed safely. The method employs a database of adjusted images to improve color accuracy and image quality. Through rigorous testing with 25 observers, the study demonstrates that the proposed approach significantly outperforms traditional editing. The performance can further enhance with an expanding database of artworks, though limitations remain.
E N D
An Exemplar-Based Method for Automatic Visual Editing and Retouching of Fine Art Reproduction Jun Jiang and JinweiGu 11/07/2013
What Is Fine Art Calligraphy (2D) Painting (2D) Sculpture (3D) Photography (2D)
What Is Fine Art Reproduction Capture Display Artwork Print Smartphone Pad Goal: make the fine art reproduction an accuraterepresentation of the original
Why Fine Art Reproduction Important • Due to safety and space limit, many collections are rarely exhibited • Too delicate to handle Munich Agreement Tangibook Leafsnap
Widely Varying Image Quality • Milkmaid (http://www.artstor.org) [Frey and Farnand. 2011]
Challenge Can we avoid or alleviate visual editing and retouching?
Visual Editing and Retouching D65 D50
Image Adjustment Functions Contrast • Global • Local Sharpness Color selected for adjustment Hue L* Chroma Threshold
Exemplar-Based Method D65 D50 Database Exemplar-based model
Distance Metric • Learn distance metric from adjusted images Image difference, RMSE, SSIM, Color difference, … Difference in the space of adjustment parameters
Visual Editing (Global Adjustment) • With a new painting • Apply visual editing Distance in the space of adjustment parameters between the new image and the existing images in the database
Retouching (Local Adjustment) • Dominant colors are extracted by K-means
Data Acquisition • Reproductions of 15 paintings were captured in a diffuse lighting geometry using a Canon 60D • 25 Observers participated the experiment making visual editing
Improvement Original Global Local Sharpness
Conclusions • We proposed an exemplar-based method to automate visual editing and retouching • The proposed method is significantly better than the average of the visual editing by observers • Its performance can be improved as the database grows • Limitations