Content Based Image Retrieval by Using Color Descriptor and Discrete Wavelet Transform

Area of Science:

• Computer Science
• Information Technology
• Digital Image Processing

Background:

• Multimedia complexity has increased, making similar content retrieval a challenge.
• Content-Based Image Retrieval (CBIR) uses visual features for image searching.
• Effective feature extraction is crucial for CBIR performance.

Purpose of the Study:

• To propose an improved CBIR method for accurate multimedia content retrieval.
• To explore the effectiveness of combining different feature extraction techniques.
• To evaluate the proposed method's performance against existing approaches.

Main Methods:

• Utilized YCbCr color space for image representation.
• Implemented canny edge histogram and discrete wavelet transform for feature extraction.
• Employed Artificial Neural Networks (ANN) for image classification and retrieval.
• Tested the algorithm on the standard Wang image database.

Main Results:

• The combined approach of edge histogram and discrete wavelet transform enhanced retrieval performance.
• Different wavelet functions were compared to determine optimal suitability.
• The proposed method demonstrated superior performance compared to existing CBIR techniques.
• Evaluated using Precision and Recall metrics.

Conclusions:

• The proposed CBIR method, integrating YCbCr color, canny edge histogram, and DWT, offers significant improvements.
• This approach provides a more effective solution for content-based image search.
• The study highlights the importance of combined feature descriptors for robust image retrieval.