Abstract:
This study focuses on advanced techniques for visual feature extraction in digital images, particularly for improving the performance of Content-Based Image Retrieval (CBIR) systems. The proposed approach utilizes the HSV (Hue, Saturation, Value) color space to extract meaningful color features from images. In this method, the HSV color space is quantized into non-uniform intervals, and a one-dimensional feature vector is constructed. Color information is then represented using a cumulative histogram, which effectively captures the distribution of colors within an image.
In addition to color analysis, texture feature extraction is performed using the Gray-Level Co-occurrence Matrix (GLCM) and the Color Co-occurrence Matrix (CCM). These methods analyze spatial relationships between pixels to capture texture patterns present in the image. By combining HSV-based color features with texture information obtained from GLCM and CCM, a multi-feature representation of images is created.
The image retrieval process is implemented using a normalized Euclidean distance classifier, which measures similarity between image feature vectors. Experimental results demonstrate that integrating color features with texture features derived from the Color Co-occurrence Matrix (CCM) significantly improves retrieval accuracy compared to traditional single-feature methods.
Color Correlogram for Image Retrieval
A color correlogram is an effective technique used in Content-Based Image Retrieval systems to describe both the distribution of colors and the spatial relationships between different colors in an image. Unlike simple color histograms, a color correlogram provides information about how pairs of colors occur at certain distances within the image.
Color features are important because they provide valuable information about the surface properties of objects in an image. However, color appearance can change depending on several factors, including lighting conditions, surface orientation, and camera viewing angles. Therefore, advanced techniques such as color correlograms help capture spatial color relationships more accurately, improving image retrieval performance.
Image Retrieval
An image retrieval system is a computer-based system designed to search, browse, and retrieve images from a large database of digital images. Traditional image retrieval techniques rely on metadata, such as captions, keywords, or textual descriptions associated with images. In these methods, images are retrieved based on matching the search query with the annotated metadata.
However, manual image annotation is time-consuming, labor-intensive, and costly, especially when dealing with large image databases. To overcome these limitations, researchers have developed automatic image annotation and content-based retrieval techniques that analyze visual features such as color, texture, and shape directly from the images.
With the rapid growth of social media platforms and web-based applications, there has been increasing interest in automated and web-based image annotation tools. These systems enable efficient organization and retrieval of images in large digital collections.
One of the earliest microcomputer-based image database retrieval systems was developed at the Massachusetts Institute of Technology during the 1980s by researchers including Banireddy Prasaad, Amar Gupta, Hoo-min Toong, and Stuart Madnick. Their work laid the foundation for modern image retrieval systems used in many applications today.