Plain, edge, texture (PET) block classifier using Tchebichef moments and SVM

Abstract

This paper presents an image block classification method using Tchebichef moments (TMs) and support vector machine (SVM). The test images are divided into non-overlapping 16 × 16 blocks and transformed into moment domain using Discrete Tchebichef Transform. These moment features are then used in the image content (block) classification. SVM is used for learning and classifying the blocks into three types: "plain", "edge" and "texture", based on their moment energy level. Experimental results show that the proposed method works well and the classification accuracy is 98.7%.

Plain, Edge, and Texture Detection Based on Orthogonal Moment

“…The reported results reveal that the PCTR [2] attains the recognition rate of 88.32% which is the minimum recognition level. While the PCSV [24] shows better recognition rate than PCTR with an improvement of 2.54%. On the The performance of the presented algorithm and existing works was determined in terms of computational time.…”

“…The dataset of the existing works is very small. For example, [24] has selected manually 150 images for each pattern. However, the dataset is considered small to predict the various angles and cases for edge pattern.…”

“…In the presented algorithm, we utilize SVM as a classifier with the RBF as a kernel. The comparison is performed with two previous studies: image pattern classification based on threshold rules (PCTR) [2] and image pattern classification based on SVM (PCSV) [24]. The proposed work is compared with PCTR and PCSV because these works are considered the most related existing works that need more insight to be more accurate and faster in the image pattern classification.…”

“…These moments are based on their polynomials and extract the local features from the region of interest of an image that has unique contents (such as edges and texture). Different feature extraction classifiers have been developed using machine learning such as Support Vector Machine (SVM) which produces an accurate PET classification [24]. Typically, in image processing applications, the image's frame is partitioned into blocks to be processed separately.…”

“…PET classification provides good results even though the threshold values for the classification process are determined from empirical results. To overcome this issue, Tchebichef moment has been used in the image block classification based on SVM by dividing the image into nonoverlapping blocks, then transforming these blocks into the moment domain to extract features that are used in the image content classification [24]. The classification has been carried out using SVM to classify the image blocks into PET blocks (plain, edge, and texture) based on the level of moment energy with an accuracy equal to 98.7%.…”