Image Retrieval Using the Fused Perceptual Color Histogram

Area of Science:

• Computer Vision
• Image Processing
• Cognitive Science

Background:

• Mimicking human cognition for visual feature extraction in image retrieval is challenging.
• Opponent color and HSV color spaces are effective in simulating human visual perception.

Purpose of the Study:

• To improve and extend the CDH (Color Difference Histogram) method using a multi-stage model.
• To develop a visual feature descriptor that mimics human perception by incorporating spatial layout, intensity, edge orientation, and opponent colors.
• To enhance the CDH distance formula for more effective image similarity adjustment.

Main Methods:

• A novel visual feature descriptor combining histogram advantages with perceptual factors.
• An improved distance formula for the CDH method with adjustable parameters.
• A multi-stage model for image feature extraction and representation.

Main Results:

• The proposed method effectively describes color, texture, and spatial features.
• Experimental results on four benchmark datasets show superior performance compared to existing methods like Color Volume Histogram, Color Difference Histogram, Local Binary Pattern Histogram, Multi-Texton Histogram, and SURF-based approaches.
• The method demonstrates efficient performance in similar image retrieval.

Conclusions:

• The developed visual feature descriptor and improved CDH method offer a more human-like approach to image representation.
• The proposed technique significantly enhances the accuracy and efficiency of similar image retrieval tasks.
• This work contributes a robust method for image retrieval by effectively integrating perceptual factors into feature extraction.