Enhancing Texture-Based Image Retrieval using GLCM and DBSCAN on a Multifaceted Dataset

Texture based image retrieval is an important aspect of various computer vision applications. In our research, we have proposed a method to enhance texture based image retrieval by utilizing Gray Level Co-occurrence Matrix (GLCM) feature extraction and Density Based Spatial Clustering of Applications with Noise (DBSCAN) clustering. To evaluate our approach, we have utilized the Corel 10k dataset, which consists of 10,000 diverse images from different categories. Our methodology involves several steps. Firstly, we convert the images to grayscale and normalize the pixel values. Then we extract four significant features from the GLCM: entropy, energy, contrast, and correlation. These features play a key role in determining image similarity. Subsequently, we apply DBSCAN clustering to refine the retrieval results based on these GLCM features. To assess the performance of our approach, we employ different distance metrics such as Euclidean, City Block, Bray Curtis and Canberra. Through experimental analysis, we have obtained promising results that highlight the effectiveness of our proposed method. The GLCM DBSCAN approach consistently outperforms using GLCM alone when it comes to retrieval precision. Among the distance metrics used for evaluation, Canberra distance achieves the highest precision values for measuring similarity between GLCM based features in the Corel 10k dataset. This indicates its suitability as a measure of similarity in this context. Overall, our research contributes to enhancing texture based image retrieval by employing GLCM feature extraction and DBSCAN clustering methods. The successful evaluation results validate the effectiveness of our approach and offer valuable insights for future improvements in this field.