While thermal optical registered datasets are becoming widely available, most of these works are based on image pairs which are pre-registered. However, thermal imagers where these images are registered by default are quite expensive. We present in this work, a thermal image registration technique which is computationally lightweight, and can be employed regardless of the resolution of the images captured. We use 2 different thermal imagers to create a completely new database and introduce it as a part of this work as well. The images captured are based on 5 different classes and encompass subjects like the Prayagraj Kumbh Mela, one of the largest public fairs in the world, captured over a period of 2 years.