Deep Learning Based Image Classification Using Multi Stage Techniques In Remote Sensing Data

Abstract

The Image enhancement process represents the most effective filtering technique for enhancement from LISS IV remotely sensed band images. Most of the common filtering techniques used are based on spatial, frequency domain filters. In particular, filters like the wiener filter, median filter, and mean filter, with selected noise removal salt, pepper, and gaussian filtering techniques. In particular, filtering with contrast stitching and brightness enhancement will apply to the image to improve the quality of image clarity with the standard techniques and normal enhancement process of the color image to increase the brightness, color, contrast, and sharpness. Hence, with an unsharp filter and histogram plot with masked image representation, the process of enhancing image quality is increased. newlineThe edge detection for most of the images is much more complex than the normal image edge detection process. In the aerial view, images able to identify the object in computer vision camera capture object able to detect the edges but in remote sensing images, a wide range of edges with a selected area to detect the edges is more complex. However, edge detection techniques are used to identify the points in an image with discontinuities. Hence, applying filters will improvise the edge in remote sensing images like road areas, river floating, and building locations in satellite image conversion of grayscale. In general, smoothing, equalized threshold, and binary with OTSU are improved. In the proposed work the sample image is into the following filters Sobel, Prewitt, Laplacian, and Canny edge detection with PSNR value being calculated to identify the most suitable filter for edge detection with canny edge detection. In addition, the value received in the filters canny produces the best results. newline newline