Classification of Noisy Image based on Statistical Feature Extraction & Designing an Adaptive Noise Removal Filter

The most important challenge in digital image processing is to de-noise the corrupted image. This project reviews existing de-noising algorithms such as AMF, GMF & HMF and their comparative study. In the proposed method first type of noise is detected by extracting the statistical features of corrupted image. In training part we are creating a knowledge base for storing the statistical features of different types of noisy images. In testing part the given query image is classified according to the statistical features by comparing the query image features to the images stored in ANN training. And then the adaptive filter is selected according to type of noise. The output of filter is again requiring tuning so for that we are using Minimum Mean Square Error filter. The result shows the comparison of existing filters with proposed filter by MSE and PSNR values so that proposed filter is efficient and effective.