Temu Kembali Citra Menggunakan Metode Local Binary Pattern Rotation Invariant (Lbprot) dan Cosine Distance Similarity

Abstract

Image retrieval is a method of searching for an image by comparing the query image with the image in the database. One of the important things in image retrieval is the feature extraction process. Currently, the feature extraction method needed is reliable in recognizing images that are rotated at various angles or invariant to rotation such as the Local Binary Pattern Rotation Invariant (LBPROT) method. In addition to the feature extraction method, what is also important is the distance measurement method. The distance measurement method used is the Cosine Distance Similarity method. the combination of these two methods resulted in the highest average precision and recall at a ratio of 70%:30% of 98.06% and 96.78% for normal images and for images that were rotated at an angle of 90o of 97.08% and recall of 96,70%. In addition, testing on rotational images produces the lowest average precision and recall at an angle of 45° at 0%.