Human detection and localization is one of the importance aspects in computer vision. It has broad applications in surveillance, robotic, driver assistance system, and for the military applications. The task is difficult because it depends on various conditions such as illumination, distance, human pose and weather condition. This study aimed to investigate human detection methods for thermal and visible images. We have explored three methods which are histogram of oriented gradient, integral image and aggregate of channel features. Our result showed that histogram of oriented gradient outperformed the other two using the tested images. However, the method is only applicable when the human is on the standing or upright position and limited to a certain distance between the scene and the camera position.
This paper represents an effective method to enhance colour correction for tongue diagnosis. Colour Correction means exactly that, correcting the colour in image in the post production process. If the task of correcting an image is not in the workflow, it may be missing out on how great the image could look. It is one of the tasks of being a director of photography to get the best image to the colourist that can be. Understanding the camera’s limitation and using that knowledge on set is the first step in getting images that are much easier to correct later in the production workflow. Cameras record three colour responses of Red, Green and Blue (RGB) that are device dependent. In this work, preliminary research on tongue colour correction using polynomial regression algorithm has been implemented on Munsell colour checker for future tongue colour correction and diagnosis. The attained RGB colour space from Munsell Checker image converted to Lab colour space which is device independent colour space based on human visual system that is perceptually uniform. Then, several degrees of polynomial regression method are employed to provide comparative analysis on colour reproduction index to produce good quality of image after colour correction procedure. The experimental outcomes on colour checker show the colour difference is equal to 3.3289, ∆E*ab=3.3289 which is acceptable in digital image colour reproducibility.
Suzaimah Ramli, Tuan Khalisah Tan Zizi @ Tuan Zizi, Norulzahrah Mohd Zainudin, Nor Asiakin Hasbullah, Norshariah Abdul Wahab, Noor Afiza Mat Razali, et al.
Thermal imaging technology can be used to detect aggressive levels in humans based on the radiated heat from their face and body. Previous researches have proposed an approach to figure out human aggressive movements using HornSchunck optical flow algorithm in order to find the flow vector for all video frames but still not strong enough to confirm and verify the existence of an aggressive movement. In this work, we propose an approach by using thermal videos for frontal views of the human body which is face view. Then, video frames are collected using thermal camera and further extracted into thermal images. We use thermal imaging to monitor the face including prefrontal and periorbital region’s thermal variations and test whether it can offer a discriminative signature for detecting aggressiveness. We start by presenting an overview of 3400 thermal images extracted from 50 participants. The results obtained is promising where aggressive and non-aggressive features can be detected by using color-based approach.