Content-based Image Retrieval Using Combining PCA and LDA Methods

Nowadays, digital images are widely used in the diagnosis of disease, facial and fingerprint, security systems, and more. Therefore, providing an accurate algorithm in image recognition and retrieval is very important. This paper presents a combination of PCA and LDA methods for image retrieval. In this method, first, the color images in the RGB space are transferred to the HSV space, then the color, shape, and texture properties are extracted from the "V" component of the HSV color space. The proposed feature vector is then constructed using the LDP histogram, color histogram, Tamura histogram, and common event matrix. Then, by combining the two methods of PCA and LDA, the specificity is reduced and finally, the classification is done. Four scenarios were designed and evaluated to evaluate the proposed method. According to experimental result and evaluation criteria, The accuracy obtained was 97.6 which indicates the proper performance of the proposed method compared to similar tasks.