Multiclass Noisy Image Classification Based on Optimal Threshold and Neighboring Window Denoising

Classification of multi class images is very enviable for different recognition. This is affected by many factors such as noise, blur, low illumination, complex background, occlusion etc. Noise is one of the major factors causing degradation of the classification performance. This paper proposes an efficient method for classification of multi class object images which are corrupted by Gaussian noise. A wavelet transform based denoising scheme by thresholding the wavelet coefficients namely NeighShrink has been utilized to eliminate too many wavelet coefficients that might contain noise information and selecting useful coefficients. This work shows robustness of proposed method of multiclass object classification over the spatial domain denoising and feature extraction method for classification.