News

Two research  findings of the USC Media Communications Lab were presented at the IEEE International Conference on Consumer Electronics (ICCE), which was held from January 9th to 12th 2015, in Las Vegas .

Auto exposure (AE) control is an important function of modern digital cameras. Simple AE algorithms are designed for a specific type of camera sensors. Advanced AE techniques have been developed to tackle a wider class of camera sensors and/or high contrast lighting conditions, yet they are computationally intensive and, thus, difficult to implement in a resource-constrained environment such as phone cameras. Besides, none of existing solutions provides robust performance if erroneous exposure occurs. To address the aforementioned shortcomings, a fast and robust AE algorithm is  in need. The first paper entitled with “Fast and Robust Camera’s Auto Exposure Control Using Convex or Concave Model”, co-authored by Yuanhang Su and C.-C. Jay Kuo, proposed a concave/convex function model for the luminance characteristics of a camera. Based on such a model, a proper parameter value can be computed using a modified secant algorithm with fast convergence. This paper was presented by an MCL member, Joe Lin.

The second paper entitled with “Uncalibrated Multiview Synthesis based on Epipolar Geometry Approximation”, co-authored by Young Ju Jeong, Hyoseok Hwang, Dongkyung Nam and C.-C. Jay Kuo, was mainly conducted in Samsung Advanced Institute of Technology (SAIT) with the input from Professor Kuo. This paper proposed an  efficient multiview rendering algorithm that takes uncalibrated stereo as the input. First, the epipolar geometry of multiple viewpoints is analyzed for multiview display. Then, the camera pose for an arbitrarily selected viewpoint is estimated by algebraic approximation. Finally, by exploiting rectification homographs and disparities of rectified stereo, one can determine multiview images with [...]

Professor C.-C. Jay Kuo attended 2015 APSIPA Annual Summit Conference (ASC) in Siem Reap, Cambodia, December 9-12. APSIPA represents “Asia Pacific Signal and Information Processing Association”, which was founded in 2009. Professor Sadaoki Furui was the first President (2009-2012) while Prof. Jay Kuo was the second (2012-2014). During his tenure, Professor Kuo launched a major membership campaign through recruiting APSIPA Friend Labs. Right now, APSIPA has around 3200 e-members in the LinkedIn Social Network and 260 APSIPA Friend Labs.

Professor Kosin Chamnongthai of King Mongkut’s University of Technology (Thailand), Professor Hitoshi Kiya of Tokyo Metropolitan University (Japan) and Professor Kuo were General Co-Chairs of APSIPA ASC 2014. The conference location, Siem Reap, is the home to the incredible Angkor ruins, a sprawling World Heritage-listed complex of ancient temples, including Angkor Wat, Angkor Thom, etc. Professor Kuo met a couple of MCL alumni from Korea and Taiwan.

Professor Kuo completed his two-year term of APSIPA President immediately after this conference, and Dr. Haizhou Li of Singapore has succeed Professor Kuo to become the new President of APSIPA. Dr. Li is a world renowned expert and technical leader in speech processing and recognition.

Congratulations to Hyunsuk Ko, an MCL member, for passing his defense this afternoon. His thesis title is “Advanced Techniques for Stereoscopic image rectification and quality assessment”. His thesis guidance committee includes Jay Kuo (Chair), Sandy Sawchuk and Aiichiro Nakano (Outside Member). The committee gave a lot of praise to the quality of Hyunsuk’s thesis and his excellent presentation. The following is the abstract of Hyunsuk’s thesis.
“New frameworks for an objective quality evaluation and an image rectification of stereoscopic image pairs are presented in this work. First, quality assessment of stereoscopic image pairs is more complicated than that for 2D images since it is a multi-dimensional problem where the quality is affected by distortion types as well as the relation between the left and right views such as different types/levels of distortion in two views. In our work, we first introduce a novel formula-based metric that provide better results than several existing methods. However, the formula-based metric still has its limitation. For further improvement, we propose a parallel boosting system based quality index. That is, we classify distortion types into groups and design a set of scorer to handle them separately. At stage 1, each scorer generates its own score for a specific distortion type. At stage 2, all intermediate scores are fused to predict the final quality index with nonlinear regression. Experimental results demonstrate that the proposed quality index outperforms most of state-of-the art quality assessment methods by a significant margin over different databases. Secondly, a novel algorithm for uncalibrated stereo image-pair rectification under the constraint of geometric distortion, called USR-CGD, is presented in this work. Although it is straightforward to define a rectifying transformation (or homography) given the epipolar geometry, many existing [...]

In November 2014, a new visiting scholar, Dr. Zhaojun Nie joined MCLab. Dr. Nie received her PhD degree from McMaster University at Hamilton, Canada in June 2014. Now we have an interview with her, talking about her research interests and future expectations.

1. Could you briefly introduce yourself and your previous research experiences?

I received my BS degree in Electrical Engineering from Xi’an Institute of Post & Telecommunication in 2007, and received my MSc degree in Optoelectronics Engineering from Beijing Institute of Technology in 2009. In June 2014, I received the PhD degree in Biomedical Engineering from McMaster University at Hamilton, Canada. My previous research focused on optical imaging and spectroscopic system design for biomedical and clinical applications, as well as clinical data analysis and classification to increase the diagnostic accuracy of tumor detection.

2. What is your first impression of USC and MCLab?

USC is a famous university. I am glad that I have this opportunity to visit USC and join the MCLab. USC campus is very beautiful. The students are all very smart and hard-working. The MCLab is a big family, where members frequently exchange ideas, share experiences and help each other. The atmosphere of active discussion is quite inspiring. Prof. Kuo is also very nice and students benefit from his prompt feedback and guidance. In addition, I think the arrangement of weekly seminar is very helpful, which provides me with a chance to know everyone’s research work. In general, I enjoy my time at MCLab very much.

3. What is your future expectation in MCLab?

The MCLab is known by its researches in image processing and computer vision. Some related techniques are widely used in various fields, including my previous research focus, medical and clinical applications. Therefore, [...]

Content-aware image retargeting is a technique that resizes images for optimum display on devices with different resolutions and aspect ratios. Traditional objective quality of experience (QoE) assessment methods are not applicable to retargeted images because the size of a retargeted image is different from its source. Dr. Jiangyang Zhang, a former MCL member and Professor C.-C. Jay Kuo, MCL Director, identified three main determining factors for humans visual QoE on retargeted images. They are global structural distortion (G), local region distortion (L) and loss of salient information (S). Zhang and Kuo selected features to quantify these respective distortion degrees and developed objective quality assessment index, called GLS, to predict viewers’ QoE by fusing selected features into one single quality score. The proposed GLS quality index has stronger correlation with human QoE than other existing objective metrics in retargeted image quality assessment with respect to two subjective image retargeting quality databases. The work was presented in the ACM Multimedia Conference on November 5 in Orlando, Florida.

A joint photo of Dr. Zhang and Prof. Kuo and a photo of Prof. Kuo together with ACM MM conference organizers and a Keynote Speaker, Prof. Rosalind Picard of MIT Media Lab (number 3 from the right), are shown.