Central AsianResearch Index

HYBRID OPTIMIZATION AND DEEP LEARNING BASED COLOR CLASSIFICATION FOR FRACTAL PATTERN GENERATION IN DIGITAL DESIGN

Fractal-based methods are used to create geometric patterns for textile and carpet design. Color selection is constrained by production cost, composition, and visual quality. A computational method is proposed for color generation, classification, and optimization. Global search is performed by a genetic algorithm. Local refinement is conducted by the simplex method. Color combinations are classified by a convolutional neural network. Production cost and color quality are balanced through a weighted objective function. Optimization weights are adjusted through user feedback. Design priorities are therefore modified during each iteration. The method is evaluated with fractals derived from Pascal’s triangle. Multiple candidate solutions are produced and locally refined. Patterns are assignedtowarm,cool,orhigh-contrastcategories. AnAUCof1.00isobtainedundertheexperimental conditions. A labor cost reduction of 10–15% is reported for the evaluated setting. Automated color selection for digital carpet design is supported by the method.

Сведения

Опубликовано
2026 · Т. 14, № 3, С. 127-140
Язык
Английский
ISSN
2306-6172 · 2308-9822
Издатель
L. N. Gumilyov Eurasian National University
Аффилиации
Television-Media Technologies
Obuda University,
IT Center under NASP
J. Selye University
John von Neumann University