Integration of Basic Descriptors for Image Retrieval

Several methods have been proposed to retrieve images from the database based on descriptors such as color, texture, and shapes. Most of the retrieval methods considered only individual descriptor to retrieve image from database which does not provide more accuracy. The large volume of images is generating day by day from the Internet of things (IoT) devices. To provide better retrieval accuracy of image from large database integration of the multiple descriptors are required. In this work, two approaches, additive and sequential are used to integrate multiple descriptors and the performance has been evaluated to enhance the accuracy of image retrieval. To analyse and establish the result, both approaches individual and integration of multiple descritpors have been evaluated with real-world databases and using various image retrieval techniques as a means for evaluation. This has been done by extracting features of the image; techniques used are color histogram, discrete wavelet transform, and canny edge detector for color, texture, and shape, respectively. The detailed analysis of individual and multiple approaches shows that several combinations of the color, texture, and shape features enhance the retrieval accuracy. The experimental results demonstrate the effectiveness of the method used for analysis.