Using Deep Learning Neural Networks to Enhance HDR Images

Authors

  • Ahmed Hasan khanjar Computer Science Faculty of Basic Education, Mustansiriya University Baghdad, Iraq

DOI:

https://doi.org/10.31695/IJERAT.2023.9.5.1

Keywords:

Deep Learning Neural Networks, HDR images, light intensity

Abstract

In recent days, photography becomes main in our daily life style. Everyone like to upload images on facebook, instagram, or any social media site, or maybe want to photograph some personal images or for memory.

In other hand, the development of technology, especially in cell phones, let these mobiles equipped with cameras with a specific resolution. We define cameras with two parameters. First one is the width and the high of the image which can define the number of pixels inside the image, the bigger these pixels can be the accuracy and true scene can get. The second in the resolution of the pixel which means how many bits are used to define the colors of the image.

All natural images have a wide dynamic range because of containing a large range of intensity values. This high range cannot be collected using non-HDR cameras. So, mobile cameras might give very bad resolutions.

This search aims to use conventional neural networks to create HDR images using a programmed manner so we can solve the problem of poor cameras problems, first of all we will create a map for an image using a dataset of static images has a different light intensities for the same scene (using differently exposed LDR images), and in the next task we will use the trained model to get the optimal tuning for the image to result in the right color space. Here we could select a group of main parameters which are the number of learning tasks, the learning strategies, the sensor data used, number of input exposures.

Downloads

Published

2023-05-28

How to Cite

Using Deep Learning Neural Networks to Enhance HDR Images. (2023). International Journal of Engineering Research and Advanced Technology (ijerat) (E-ISSN 2454-6135) , 9(5), 1-11. https://doi.org/10.31695/IJERAT.2023.9.5.1