Reconstructing the Medical Image by Autoencoder with Stochastic Processing in Neural Network

The working on image to perform different operation is known as image processing. Digital Image is a collection of a defined number of pixels. Each feature has a particular pixel value at a specific position. Image restoration can be considered as an image enrichment where the final image is printed with more important features that create more appealing for the spectator but it is not essential to create genuine data from the point of a scientific vision. Autoencoder is one of the techniques in artificial neural network to analyze and predict a huge volume of data with different many features. This technique has three stages namely the first one is encoding, second one is compression and third one is decoding. Denoising autoencoder is an enhanced technique for image processing to reconstruct the damaged image to original image. Denoising autoencoder techniques are applying on the medical image to get the high quality image in digital image processing. In this paper, stochastic gradient descent algorithm is applied to predict the damaged pixel and replaced by new pixel values. The result is better than previous result in medical images.