A DWPT domain transform and COM statistics method combined with rough set for images classification
- PDF / 895,354 Bytes
- 20 Pages / 439.37 x 666.142 pts Page_size
- 70 Downloads / 173 Views
A DWPT domain transform and COM statistics method combined with rough set for images classification Ching-Hsue Cheng 1 Received: 21 January 2020 / Revised: 24 July 2020 / Accepted: 31 July 2020 # Springer Science+Business Media, LLC, part of Springer Nature 2020
Abstract
This paper compares the discrete wavelet packet transform combined with rough sets (DWPT-RS) and co-occurrence matrix combined with rough sets (COM-RS) for invariant pixel region image classification. In verification, the Brodatz and building material datasets were employed in experiments to verify the performances of the two methods. Moreover, the experiments have been performed recursively for over 600,000 rounds, and employed different level DWPT, different wavelet filter, and attribute data discretization to find the best results. The results show: (1) the proposed method outperforms the listing methods in accuracy, (2) the non-discretization classification accuracy is better than the discretization classification accuracy, (3) the 2-level DWPT is better than that of the 1-level DWPT, (4) the sym3 wavelet filter has better performance than the listing filters, and (5) the proposed DWPT-RS method has a fast computation time than the listing methods. The results can help people who want to perform related research in the future to choose the best method to match it more effectively. Keywords Rough sets (RS) . Discrete wavelet packet transform (DWPT) . Co-occurrence matrix (COM) . Image classification
1 Introduction Image analysis is the extraction of meaningful information from images, and texture analysis is a branch of image analysis. Many researchers have studied in the field of texture analysis during the last decades [5, 9, 18, 19, 21, 32, 40]. The goals of texture analysis are to understand the model and process texture in computer vision and to simulate the human visual learning
* Ching-Hsue Cheng [email protected]
1
Department of Information Management, National Yunlin University of Science and Technology, Douliou, Yunlin, Taiwan
Multimedia Tools and Applications
process. Texture classification can systematically organize image data into more readily interpretable information, which is used in a wide range of applications such as industrial inspection, image retrieval, medical imaging, and remote sensing. Based on the rapid development of computer vision technology, camera vision inspection has received increasing attention. From the perspective of algorithms, the visual inspection methods can be divided into four categories [40] as follows. (1) Statistical methods: These methods use spatial information to evaluate textures that are generally applicable to irregularly textured surfaces. The histograms of gray histogram and color histogram, co-occurrence matrix, local binary mode (LBP), and invariant rotation measure of local variance (RIMLV) are usually used as an effective function. (2) Structure description and analysis: The methods are to extract features, such as regularity and local orientation are extracted from the sp
Data Loading...
