Unique Image Representation as a Tensor

Author

Bharat Rosanlall, CUNY City CollegeFollow

Date of Award

2020

Document Type

Thesis

Department

Computer Science

First Advisor

Izidor Gertner

Keywords

Orthonormal Transform, Tensor, Coefficients

Abstract

This thesis presents a two dimensional orthonormal transform that represents an image as coefficients in 4 independent channels. The salient feature of these coefficients is that they contain complete position spatial frequency information about the image, in a sense that the original image can be reconstructed from these coefficients with negligible error. These coefficients can be used in various machine learning, AI , and other tasks where data features are used. Popular convolutional layer used in various neural networks reduces information and can not reconstruct original image. In this thesis , we present several examples where these coefficients are used in image classification tasks for a standard data set.

Recommended Citation

Rosanlall, Bharat, "Unique Image Representation as a Tensor" (2020). CUNY Academic Works.
https://academicworks.cuny.edu/cc_etds_theses/924