Multi-scale Xception based depthwise separable convolution for single image super-resolution.

The main target of Single image super-resolution is to recover high-quality or high-resolution image from degraded version of low-quality or low-resolution image. Recently, deep learning-based approaches have achieved significant performance in image super-resolution tasks. However, existing approaches related with image super-resolution fail to use the features information of low-resolution images as well as do not recover the hierarchical features for the final reconstruction purpose. In this research work, we have proposed a new architecture inspired by ResNet and Xception networks, which enable a significant drop in the number of network parameters and improve the processing speed to obtain the SR results. We are compared our proposed algorithm with existing state-of-the-art algorithms and confirmed the great ability to construct HR images with fine, rich, and sharp texture details as well as edges. The experimental results validate that our proposed approach has robust performance compared to other popular techniques related to accuracy, speed, and visual quality.

Non-Contact Respiration Monitoring and Body Movements Detection for Sleep Using Thermal Imaging.

Monitoring of respiration and body movements during sleep is a part of screening sleep disorders related to health status. Nowadays, thermal-based methods are presented to monitor the sleeping person without any sensors attached to the body to protect privacy. A non-contact respiration monitoring based on thermal videos requires visible facial landmarks like nostril and mouth. The limitation of these techniques is the failure of face detection while sleeping with a fixed camera position. This study presents the non-contact respiration monitoring approach that does not require facial landmark visibility under the natural sleep environment, which implies an uncontrolled sleep posture, darkness, and subjects covered with a blanket. The automatic region of interest (ROI) extraction by temperature detection and breathing motion detection is based on image processing integrated to obtain the respiration signals. A signal processing technique was used to estimate respiration and body movements information from a sequence of thermal video. The proposed approach has been tested on 16 volunteers, for which video recordings were carried out by themselves. The participants were also asked to wear the Go Direct respiratory belt for capturing reference data. The result revealed that our proposed measuring respiratory rate obtains root mean square error (RMSE) of 1.82±0.75 bpm. The advantage of this approach lies in its simplicity and accessibility to serve users who require monitoring the respiration during sleep without direct contact by themselves.

Application of the real-time Retinex image enhancement for endoscopic images.

This paper presents a real-time image enhancement technique for gastric endoscopy, which is based on the variational approach of the Retinex theory. In order to efficiently reduce the computational cost required for image enhancement, processing layers and repeat counts of iterations are determined in accordance with software evaluation result, and as for processing architecture, the pipelining architecture can handle high resolution pictures in real-time. To show its potential, performance comparison between with and without the proposed image enhancement technique is shown using several video images obtained by endoscopy for different parts of digestive organ.

A speed-up scheme based on multiple-instance pruning for pedestrian detection using a support vector machine.

In pedestrian detection, as sophisticated feature descriptors are used for improving detection accuracy, its processing speed becomes a critical issue. In this paper, we propose a novel speed-up scheme based on multiple-instance pruning (MIP), one of the soft cascade methods, to enhance the processing speed of support vector machine (SVM) classifiers. Our scheme mainly consists of three steps. First, we regularly split an SVM classifier into multiple parts and build a cascade structure using them. Next, we rearrange the cascade structure for enhancing the rejection rate, and then train the rejection threshold of each stage composing the cascade structure using the MIP. To verify the validity of our scheme, we apply it to a pedestrian classifier using co-occurrence histograms of oriented gradients trained by an SVM, and experimental results show that the processing time for classification of the proposed scheme is as low as one-hundredth of the original classifier without sacrificing detection accuracy.