FOOD TEXTURE DESCRIPTORS BASED ON FRACTAL AND LOCAL GRADIENT INFORMATION

Area of Science:

• Computer Vision
• Image Analysis
• Food Science

Background:

• Dietary assessment relies on accurate food identification.
• Current image analysis methods for food categorization have limitations.

Purpose of the Study:

• To develop and evaluate novel texture descriptors for food image classification.
• To improve the accuracy of dietary assessment through enhanced image analysis.

Main Methods:

• Introduced three texture descriptors: entropy-based fractal dimension (EFD), Gabor-based fractal dimension (GFD), and gradient orientation spatial distribution (GOSDM).
• Evaluated methods on the Brodatz texture database and a custom food dataset.
• Compared performance against established texture and object categorization techniques.

Main Results:

• The proposed EFD, GFD, and GOSDM methods demonstrated superior performance in food categorization tasks.
• These novel descriptors consistently outperformed widely used techniques.
• Effectiveness validated across diverse textures in both general and food-specific datasets.

Conclusions:

• The developed texture descriptors offer a significant advancement in food image analysis.
• These methods provide a more accurate foundation for automated dietary assessment systems.
• The study highlights the potential of multifractal analysis and gradient orientation features for food recognition.