Pattern Generation and Dimensionality Reduction using CNN for Image Restoration

In recent days the image retrieval process is used in many applications are medical images, image classification and image segmentation process.an image retrieval is the process of retrieving images from the database. The retrieved images are classified by numerous classification techniques. In this work, the preprocessing stage, the input images are isolated from background and foreground images. A novel window based pattern generation technique is used to generate the patterns of preprocessed image. These patterns are provide the necessary information which are present in the preprocessed image. From this pattern the effective features are extracted and these features are classified by using a novel dimensionality reduction based classification technique of convolution neural network (CNN). The proposed method performance is high compare than other classification techniques of SVM and KNN.