Trung Dung Do

Real-Time Human Action Recognition Using CNN Over Temporal Images for Static Video Surveillance Cameras

This paper proposes a real-time human action recognition approach to static video surveillance systems. This approach predicts human actions using temporal images and convolutional neural networks CNN. CNN is a type of deep learning model that can automatically learn features from training videos. Although the state-of-the-art methods have shown high accuracy, they consume a lot of computational resources. Another problem is that many methods assume that exact knowledge of human positions. Moreover, most of the current methods build complex handcrafted features for specific classifiers. Therefore, these kinds of methods are difficult to apply in real-world applications. In this paper, a novel CNN model based on temporal images and a hierarchical action structure is developed for real-time human action recognition. The hierarchical action structure includes three levels: action layer, motion layer, and posture layer. The top layer represents subtle actions; the bottom layer represents posture. Each layer contains one CNN, which means that this model has three CNNs working together; layers are combined to represent many different kinds of action with a large degree of freedom. The developed approach was implemented and achieved superior performance for the ICVL action dataset; the algorithm can run at around 20 frames per second.

Inter-level and intra-level deconvolution based image deblurring algorithm for wide field microscopy

This paper proposes an inter-level and intra-level deconvolution based image deblurring algorithm (ILILD) for microscopic images. Pyramid structure is used, and inter-level deconvolution is applied to estimate latent image from coarse level to fine level. The inter-level algorithm is based on total variation regularized Richardson-Lucy scheme, which can estimate latent image with artifacts suppressed. After inter-level deconvolution, intra-level deconvolution is applied. In each pyramid level of image, the residual deconvolution is done as the intra-level deconvolution scheme to recover image edges and details furtherly. Experiments show that ILILD algorithm can estimate latent images in less time and the results have better peak signal to noise ratio, higher image entropies and few artifacts.

Blind image deconvolution based on robust stable edge prediction

This paper proposes a robust blind deconvolution method for removing a uniform blur from microscopy images. For the estimation of the kernel - point spread function (PSF) - the stable edge is estimated using a fuzzy edge prediction method. Based on the estimated stable edges, optimizing a blurring objective function leads to a closed form for the estimation of a kernel and latent image. In comparison with existing deconvolution methods based on iterative optimization, the proposed method with a closed-form solution produces a significant decrease in processing time, which is an important barrier in applying deconvolution methods to real-world applications. Experimental results demonstrate the robustness and high efficiency of the proposed method for diverse microscopy images.

Improvement of Accuracy for Human Action Recognition by Histogram of Changing Points and Average Speed Descriptors

Human action recognition has become an important research topic in computer vision area recently due to many applications in the real world, such as video surveillance, video retrieval, video analysis, and human-computer interaction. The goal of this paper is to evaluate descriptors which have recently been used in action recognition, namely Histogram of Oriented Gradient (HOG) and Histogram of Optical Flow (HOF). This paper also proposes new descriptors to represent the change of points within each part of a human body, caused by actions named as Histogram of Changing Points (HCP) and so-called Average Speed (AS) which measures the average speed of actions. The descriptors are combined to build a strong descriptor to represent human actions by modeling the information about appearance, local motion, and changes on each part of the body, as well as motion speed. The effectiveness of these new descriptors is evaluated in the experiments on KTH and Hollywood datasets. Category: Smart and intelligent computing

An Efficient Pedestrian Detection Approach Using a Novel Split Function of Hough Forests

In pedestrian detection applications, one of the most popular frameworks that has received extensive attention in recent years is widely known as a ‘Hough forest’ (HF). To improve the accuracy of detection, this paper proposes a novel split function to exploit the statistical information of the training set stored in each node during the construction of the forest. The proposed split function makes the trees in the forest more robust to noise and illumination changes. Moreover, the errors of each stage in the training forest are minimized using a global loss function to support trees to track harder training samples. After having the forest trained, the standard HF detector follows up to search for and localize instances in the image. Experimental results showed that the detection performance of the proposed framework was improved significantly with respect to the standard HF and alternating decision forest (ADF) in some public datasets.

Full weighting Hough Forests for object detection

Object detection plays an important role in autonomous video surveillance systems nowadays. Models based on the Hough Forests are widely applied, which use the local patches that vote for the object centers in images. Since these patches vote independently from each other, there is no guarantee that trees built in Hough Forests can obtain optimal parameters for the entire model. This paper proposes a novel method to improve the Hough Forests by introducing weights to each offset in the positive training images to specify the importance of the patch to the training object. Also, all patches in the dataset are weighted and updated during the training process by minimizing the global loss function. The weights are used in both the training and detection phases to obtain a more accurate location of instances in detection images. The proposed method is then evaluated on TUD pedestrian and UIUC car datasets showing promising results compared to recent methods such as Hough Forests, and Alternating Decision Forests.