Feature Vector Generation for Content Based Image Retrieval

The utilization of computer vision techniques to address the image retrieval problem, which involves searching for digital images in extensive databases, is referred to as Content Based Image Retrieval. The present paper expounds on a CBIR approach that is capable of filtering images on the basis of their content, thereby enhancing the indexing process and yielding outcomes that are more precise. The given paper involves the generation of feature vectors through the utilization of the color averaging technique, followed by the application of similarity measures. Furthermore, the performance of the proposed method is evaluated through the selection of a few query images per class in a random manner. The Precision-Recall cross-over plot is utilized as a performance evaluation metric to assess the efficacy of the algorithm. The findings of the study indicate that the relevance of the retrieved image is influenced by both the size of the database and the number of distinct classes.