Compression of hyperspectral image using PCA–DCT technology

Abstract

Hyperspectral image is a high-resolution image containing large number of details about the presented region. To capture the hyperspectral image, different types of sensor are used in satellite such as AURA, CALIPSO, OCO, and PARASOL, whereas hyperspectral image sensors are used to measure the reflectance of each pixel value in very large amount and stored in 3D cube format. As this type of images is of large data, they possess large amount of storage size and transmission time is also much as compare to other image. Hence, compression of such image is very important to reduce the size and transmission time. Different types of compression methods had been performed on this type of image to compress and to reduce storage size and time. To overcome such problem, a unique compression method should be there to compress the image with less data loss and low transmission period. We introduced a method known as PCA–DCT method for compression which means principal component analysis followed by discrete cosine transform. We are proposing this method in combination pattern to reduce the data dimensionality without loss of data, but as DCT is lossy type of compression, we are going to use DCT to just find the coefficient of the image. The popular known Reed–Xiaoli algorithm is used for the identification of the current pixels with the background image. The quality of image is going to be calculated using the mean square error and peak signal-to-noise ratio value.