A Novel Method for the Compression of Gigapixel Image Using CORDIC Algorithm

Abstract

Here it presents a novel technique for the compression of gigapixel images by mapping them to compact latent or hidden space using CORDIC algorithm and neural networks. The fundamental problem of giga pixel image analysis lies in the poor SNR present in such images. A whole-giga pixel histopathology image is split into a series of patches that are mapped to a collection of low embedding vectors. Here the encoder operation is performed along with CORDIC Algorithm, which helps in simplifying trigonometric and exponential operations. These embeddings are processed while keeping the original patches spatially organized. The resulting array represents a compressed gigapixel image.