Open Access

ARTICLE

Use of Local Region Maps on Convolutional LSTM for Single-Image HDR Reconstruction

Seungwook Oh, GyeongIk Shin, Hyunki Hong^*
1 College of Software, Chung-Ang University, Heukseok-ro 84, Dongjak-ku, Seoul, 06973, Korea
* Corresponding Author: Hyunki Hong. Email:
(This article belongs to this Special Issue: Application of Machine-Learning in Computer Vision)

Computers, Materials & Continua 2022, 71(3), 4555-4572. https://doi.org/10.32604/cmc.2022.022086

Received 27 July 2021; Accepted 18 October 2021; Issue published 14 January 2022

Abstract

Low dynamic range (LDR) images captured by consumer cameras have a limited luminance range. As the conventional method for generating high dynamic range (HDR) images involves merging multiple-exposure LDR images of the same scene (assuming a stationary scene), we introduce a learning-based model for single-image HDR reconstruction. An input LDR image is sequentially segmented into the local region maps based on the cumulative histogram of the input brightness distribution. Using the local region maps, SParam-Net estimates the parameters of an inverse tone mapping function to generate a pseudo-HDR image. We process the segmented region maps as the input sequences on long short-term memory. Finally, a fast super-resolution convolutional neural network is used for HDR image reconstruction. The proposed method was trained and tested on datasets including HDR-Real, LDR-HDR-pair, and HDR-Eye. The experimental results revealed that HDR images can be generated more reliably than using contemporary end-to-end approaches.

---

Keywords

Low dynamic range; high dynamic range; deep learning; convolutional long short-term memory; inverse tone mapping function

---