Two-layer content-based image retrieval technique for improving effectiveness

Content-based image retrieval (CBIR) is an automated process that seeks to retrieve similar/closer images from a large-scale image collection by extracting visual content from the images themselves. In general, CBIR systems consist of two main steps: 1) feature extraction and 2) feature matching. The extraction of features entails decreasing the amount of data required to describe a large set of data. Feature matching, on the other hand, is the process of comparing the extracted features from the query image to the extracted features from images in the database using a certain distance metric. Meanwhile, the extracted features of the query image are compared to those of the images in the database throughout the retrieval process, allowing each indexed image to be ranked according to its distance from the query image. This paper exploits to take the advantage from both global and local feature and hence a hybrid CBIR technique is devised which contains two layers of filtering. The first layer uses the Bag of Features (BoF) technique to compare the query image to all images in the database in order to eliminate/exclude as many dissimilar images as possible. This results in the retrieval of a number of images that are closer to the query image. The second layer aims to compare the query image to the retrieved images earned from the first layer. This is based on the extraction of texture-based and color-based features. The Local Binary Pattern (LBP) and Discrete Wavelet Transform (DWT) were used as texture features. Color features were also used from three distinct color spaces (RGB, HSV, and YCbCr). Entropy and mean of every single channel are measured. The experiments are carried out in details utilizing the widely used and well-known Corel-1 k database. In regards of precision rate, the experimental findings show that the proposed two-layer strategy outperforms existing state-of-the-art approaches, with top-10 and top-20 precision rates of 86.65% and 81%, respectively.