Irena Galić

Technology assisting the blind — Video processing based staircase detection

Real-time video processing and analysis are getting more present in various technical systems and moreover provide potential application in systems for assisting the blind. Fast processing of video frames allows us to extract useful information and detection of objects that are important for the blind people orientation and navigation, especially in unfamiliar areas. In this paper, we propose the method for staircase detection from video frames as a part of assistance system for the blind. The method is based on morphological preprocessing operations and further vertical and horizontal analysis of the frames. The experimental results obtained by implementing proposed method show its contributions in accuracy and performance.

Epicardial fat registration by local adaptive morphology-thresholding based 2D segmentation

3D heart registration has become an important issue in cardio vascular diagnosis and treatment. This is mainly due to advanced medical imaging technologies that provide significant amount of data with high precision. One of the important features of the heart that has recently drawn attention is epicardial fat (surrounds the heart), which according to some preliminary studies can be correlated well with risk prediction of various cardiovascular diseases. Consequently, automatic detection and registration of epicardial fat is considered as important task for medical doctors to include as additional feature within the already existing software for medical imaging and visualization. In this paper, we analyze heart images obtained by 4D CT technology and propose a segmentation scheme that automatically extracts epcardial fat in each 2D slice in order to perform 3D epicardial fat registration and visualization. The segmentation algorithm first enhances input image after which it performs patch based labeling and clustering of the selected features. The experimental results indicate good epicardial fat registration performance in comparison to manual segmentation obtained by the medical doctors.

Voice Quality Testing Using VoIP Applications over 4G Mobile Networks

Nowadays, many mobile applications in the telecommunication market provide Voice over IP (VoIP), video and data services. This paper presents the performance evaluation of a 4G mobile network in the city of Pristina. This was accomplished by testing two of the most popular VoIP applications, Skype and Viber. Stationary and dynamic scenarios have been considered to evaluate the voice quality and the Quality of Experience (QoE) of both applications under the same mobile network radio conditions, throughput capacity and coverage requirements. The Mean Opinion Score (MOS) method was used to evaluate the recorded speech files based on the feedback perceptions of different users. These results were found to be very useful for the continuously increasing demand on VoIP services and applications.

Skeletonization method for vessel delineation of arteriovenous malformation

Cerebral arteriovenous malformation (AVM) presents a great health threat due to its high probability of rupture that can cause severe brain damage. Image segmentation alone is not sufficient to support AVM embolization procedure. In order to successfully navigate the catheter and perform embolization, the segmented blood vessels need to be classified into feeding arteries, draining veins and the AVM nidus. For this reason we address here the AVM localization and vessel decomposition problem. We propose in this paper a novel AVM localization and vessel delineation method using ordered thinning-based skeletonization. The main focus of vessel delineation is the delineation of draining veins since it is essential for the embolization procedure. The main contribution is a graph-based method for exact extraction of draining veins which, in combination with our earlier work on AVM detection, allows the AVM decomposition into veins, arteries and the nidus (with an emphasis on the draining veins). We validate the proposed approach on blood vessel phantoms representing the veins and the AVM structure, as well as on cerebral 3D digital rotational angiography (3DRA) images before and after embolization, paired with digital subtraction angiography (DSA) images. Results on AVM delineation show high correspondence to the ground truth structures and indicate potentials for use in surgical planning.

Real-time Multiresolution Crosswalk Detection with Walk Light Recognition for the Blind

Real-time image processing and object detection techniques have a great potential to be applied in digital assistive tools for the blind and visually impaired persons. In this paper, a ...

Technology assisting the blind — routing on the staircases using wide-angle camera

Wearable digital systems for the blind persons are more often based on video frames captured using various small-sized cameras. Image analysis and processing techniques allow delivering useful information about environment. Helping the blind in their everyday movement is one of the primary tasks of such first-person vision (FPV) systems. This work extends our previous work on staircase detection to facilitate the routing of the blind on the staircases including the obstacle avoidance. Algorithm for routing is developed and tested on portable computer using small action camera to exploit the benefits of wide-angle input videos and image stabilization feature. Directed sound guidance is proposed as a solution for informing the blind about most adequate route.

Compression parameters tuning for automatic image optimization in web applications

In this paper an automatic image optimization workflow for web applications is proposed. The workflow consists of several parts: a JPEG encoder, a perceptual visual quality metric (PVQM) and a system for automatic iterative compression of the images. The system for automatic compression iteratively compresses the image and tries to determine the highest compression rate for each image, without compromising the perceptual quality of the image. To evaluate the proposed PVQM parameter used for stopping the iterative compression, a study is performed where 155 users rated random selection from 559 versions of differently compressed images from a public dataset. All images are high quality images in high resolution, which makes them suitable for use in web applications. For JPEG encoder we evaluate very popular open source software package: mozjpeg from Mozilla. The results show which values of PVQM metric should be used to obtain a desired trade-off between compression ratio and subjective loss of quality.

Influence of the input image resolution on the staircase detection

Camera based assistance systems for the blind and visually impaired are still in a preliminary stage. Image processing techniques have already been used for the obstacle detection but the real-time processing still significantly depends on the input image resolution. This paper elaborates on the problem of staircase detection for the blind and visually impaired emphasizing the input image resolution. The evaluation of successful staircase detection is made by using different input resolutions. Choosing the appropriate input resolution is crucial to achieve optimal performance and find the compromise between accuracy and processing speed. The experiment is conducted on a dataset of frames in 5 different resolutions taken from video sequences.

Comparison of different PDE-based image compression methods with respect to the choice of PDE

In recent times there is an increasing number of publications that explore potential of partial differential equations (PDEs) as image compression methods. PDEs are frequently used for interpolation problems even when data is sparse. In PDE-based image compression only a part of data is stored by various sparsification methods while in decompression stage missing data is restored. Goal of the presented paper is to explore the potential of PDEs for image compression and importance of selecting an appropriate PDE for compression and decompression. PDE-based interpolation and image compression methods are described and a number of different PDEs are evaluated for compression.

Image compression with B-tree coding algorithm enhanced by data modelling with Burrows-Wheeler transformation

The paper shows that the partial differential based compression framework, Edge Enhancing Diffusion Compression (EEDC) on high compression ratios can come close to or even be better than present compression standard -JPEG2000 thus presenting a novel method for image compression. In this paper EEDC will be enhanced by changing its data coding, i.e. Huffman coding will be changed with an entropy coder accompanied with Burrows-Wheeler transformation and context mixing. Images, graphs and tables show image compression results. The purpose of this article is to examine the effectiveness of the PDEs in image compression and to evaluate it by comparing to cosine and wavelet transform based compression methods.