Paper Title
Enhanced Constrained Least Squares Filter for Wood Texture Classification System

Abstract
Abstract - Image restoration is an evolving technique that will aid the analyzing and interpretation process of digital images in the spatial domain. Signal noise and motion blur have been the main factors that contribute to image quality degradation on digital images. The fundamental intention of image restoration is to suppress the image degradation while persevering the significant features of the image, such as texture and edge information. In this study, an improved image reconstruction technique is proposed by using Enhanced Constrained Least Squares Filter (ECLSF) algorithm to ensure effective preservation of the image features. The proposed ECLSF algorithm consists of 3 stages which includes unsharp masking of CLSF, curvelet transform filtering with cycle spinning and iterative curvelet transform filtering. Then, a pre-trained residual neural network (ResNet) model is adopted for feature extraction process. Finally, a support vector machine is applied to perform final classification. The proposed system is compared with several image restoration algorithms for comparison purposes. Experimental results demonstrated that the proposed ECLSF algorithm managed to improve the quality of test images which contributed to classification accuracy of 98% when classifying 20 classes of wood texture images. This justifies that the proposed ECLSF is more feasible in restoring degraded wood texture images compared to previous works. Keywords - Computer vision, Image processing, Pattern recognition, SVM.