Content-based image retrieval for categorized dataset by aggregating gradient and texture features

Content-based image retrieval is the process of retrieving images similar to the Query Image. The task of finding the dissimilarity among the different objects is the primary challenge in the content-based retrieval system. The challenge increases when the dissimilarity among similar objects is less. For instance, retrieving sunflowers from the dataset having different types of objects (i.e., the uncategorized dataset) like faces, rose, ball, mountains, etc. is a challenging task but what if the sunflowers are to be retrieved from the dataset having flowers only (i.e., the categorized dataset)? Such scenarios make content-based image retrieval more challenging because now the objects are very much similar to each other (categorized) and finding dissimilarity among them is a difficult task. In the state-of-the-art approaches, much work has been done on uncategorized data as compared to categorized one. The present paper proposes an image retrieval approach out of the objects of the same category as of Query Image. Gradient and texture features have been used for the retrieval of images out of the objects of the same category. The Proposed Approach has been applied to various categories of objects like flowers (roses, lily, sunflower), toys (cars, bears, frogs), dogs (Dalmatian, Pomeranian), etc. The Proposed Approach outperforms the state-of-the-art approaches by obtaining an average accuracy of 90 $$\%$$ % .