Description

In this trademark image retrieval system, the novel base retrieval is undertaken by the number of trademark images. This method is proposed to retrieve the trademark images by detecting the edges and then extract the Local and binary features of the object. The local features are mainly extracts the interior data’s of the trademark image. The data occurred by the local and global features of the test image is saved. Then the saved data is compare to the training database, hence the training object is matched to the test image. In this method, there are 4 sets of dataset is used for the retrieval process. The centroid value detection takes a major role in the local feature extraction system. By calculating this distance from centroid to the outer distance, the robustness of the axis rotation is maintained. Hence the internal feature of the object is extracted exactly from the object. But in the global feature extraction, the Zernike moment is more essential to take the Zernike polynomial calculation value. In the local feature extraction, the distance between the centroid to the outer most distance of the object is calculated. As considered the dataset image, the entire image should be the binary image. Then the every image in the data set is converted into the same resolution. For example if the test image is converted into the 250 X 250 resolution, then all the image in the test image as well as the image in the training dataset should converted into the same resolution.