A Hybrid Classifier for Classification of Rice Crop Varieties

In this paper propose a denoising method for reducing noise in digital images. An efficient RaoBlackwellized Particle Filter (RBPF) with maximum likelihood Estimation approach is used for improving the learning stage of the image structural model and guiding the particles to the most appropriate direction. It increases the efficiency of particle transitions. The proposal distribution is computed by conditionally Gaussian state space models and Rao-Blackwellized particle filtering. The discrete state of operation is identified using the continuous measurements corrupted by Gaussian noise. The analytical structure of the model is computed by the distribution of the continuous states. The posterior distribution can be approximated with a recursive, stochastic mixture of Gaussians. Rao-Blackwellized particle filtering is a combination of a particle filter (PF) and a bank of Kalman filters. The distribution of the discrete states is computed by using Particle Filters and the distribution of the continuous states are computed by using a bank of Kalman filters. The Maximum likelihood Estimation method is used for noise distribution process. The RBPF with MLE is very effective in eliminating noise. RBPF with MLE is compared with particle filter, Markov Random Field particle filter and RBPF. In this paper different performance metrics are evaluated for this type of noise removal technique. The metrics are Mean Square error, Root Mean square error, Peak Signal to Noise Ratio, Normalized absolute Error, and Normalized Cross Correlation, Mean Absolute Error and Signal to Noise Ratio. Experimental results prove that RBPF with MLE outperforms for degraded medical images. Keywords— ECG, MIT-BIH Database, Median filter, FIR filter, Gaussian filter, Butterworth filter.