Please ensure that you have turned in everything. I asked that everyone turn in a title/abstract (which everyone did), plus a representative picture, slide, or screen capture, and that you make your code available to me electronically.

name code available? picture?
Satish Puri yes
Xiaxia Yu yes yes
David Bolding yes yes
Gang Shen yes yes
Semra Kul yes yes
Sunny Shakya yes yes
Mingyuan Yan yes yes
Sanish Rai yes yes
Mark Daniel Gibbons yes yes
Chad Griffith yes yes
Tariku Ahmed yes
Sanghoon Lee yes yes
Guoliang Liu yes yes
Zainab Haydari yes yes

CSc 8270 Presentations, Fall 2011

Satish Puri,
Convolution implementation on GPU for Magnetic Resonance Imaging
I have implemented a digital signal processing algorithm in Graphics Processing Unit (GPU). To my best knowledge, MATLAB does not directly take advantage of parallel processing capabilities of a graphics processing unit. As such, it is interesting to explore GPU implementation of existing MATLAB function e.g. *conv* and *conv2* (convolution).

Xiaxia Yu,
Sparse Dictionary for HIV-1 Protease Drug Resistance Prediction
The sparse signal representation is an emerging research area in the recent years. Indeed, it has been realized that the sparsity is a hidden prior information for most of the signals in the physical world and the related phylosophy and algorithms have been applied in a wide range of areas such as signal acqusition, signal representation, pattern classification, etc., and have achieved very exciting results. Therefore, in this project, we aim at applying such technique in our project of the HIV-1 protease drug resistance prediction. Specifically, we learn the dictionaries to represent the drug resistant and non-drug resistant mutants of the HIV-1 protease. Then, given a new protease, the learned dictionaries are used to sparsely represent the new protease. The protease is then classified according to the error of the sparse representations. The proposed method is tested on 360 mutants of HIV-1 protease to demonstrate the algorithm's accuracy and robustness.
results slide from Xiaxia's presentation

David Bolding,
Applying the novel image-blurring technique from my paper summary
My presentations is about applying the novel image-blurring technique from my paper summary. In mathematical terms, it's a coordinate transform across a manifold. It's much simpler in implementation terms. I'll probably have some snide comments about unnecessarily dressing up a published result with mathematical language. I got the blur to work; they also included other filters, like a colorizer and an edge-detecter. I'll probably try to add those over the break, so I have something more to present.
figure 1 from David's work

Gang Shen,
A Study of Huffman Lossless Image Compression Algorithms
The project experimented various Huffman lossless image compression algorithms. A new Huffman lossless image compression program is also proposed and tested. Experiments show the proposed new algorithm achieved better performance than all tested compression algorithms including a published "Folding" based compression algorithm.
results slide from Gang's presentation

Semra Kul,
Segmentation of Intervertebral Discs in Spine MRI
In the project, a method for segmentation of intervertebral discs from the surrounding body parts in spine MRI (Magnetic Resonance Imaging) is presented. The morphological image processing (thresholding, erosion, dilation) are applied to sagittal (vertical) spine MRI samples. The aim of the project is to provide a preliminary work for further segmantation of the spine in diagnosis of disc related diseases.
a representative slide from Semra's presentation

Sunny Shakya,
Segmenting an Image using Snakes
Description - Snake is the widely used method for segmenting an image or detecting an object boundary. I have implemented Snake in MATLAB and Since snake has very limited capture range, an extra pressure force is added to increase its range of capture and the resulting snake is known as Balloon Snake. I have also compared my implementation with Gradient Vector Flow Snake. Experiments shows that Snake can accurately segment an image provided snake is initialized close to the region of interest.
slide from Sunny's presentation

Mingyuan Yan,
Image Denoising Using Diffusion and BayesShrink-based Wavelet Transform
Among all the wavelet based image denoising technique, wavelet image denoising based threshold filtering is the most popular due to its low computation, easily implementing and so on. In this paper, the author first introduces some important preliminary about image denoising. After that, An algorithm based on wavelet and diffusion is implemented followed its performance evaluation by comparing with four other approaches. The experiment result shows that the diffusion and wavelet-based method outperforms the traditional thresholding method and local mean based method is some cases. However, the visual result shows that the image quality after denoising is far from satisfying compared with the clean image quality.
results slide from Mingyuan's presentation

Sanish Rai,
Image stitching
Image stitching finds it use in creating a single map or photo by combining multiple small fragments. The method discussed works by matching multiple images sequentially and finding invariant local features. These features are used to find match between the images. The use of invariant local features makes the method insensitive to the ordering, orientation, scale and illumination of the multiple images.
image from Sanish's project

Mark Daniel Gibbons,
JWRMark - A Java Library for Watermarking
Watermarking has been gaining traction in recent years to be able to detect who distributed an image [9] and to ensure content is legitimate [8]. A Java library, JWRMark, is proposed to provide an easy means for Java developers to embed a watermark within an image. Two algorithms are included with the first release. The first algorithm is one which simply adds the watermark to the LL band of an image decomposed by a Discrete Wavelet Transform (DWT). The second algorithm embeds the image within the singular values of an image decomposed by the DWT and Discrete Cosine Transform (DCT). An emphasis is put on being able to easily add new algorithms with reusable code.

[8] Lintao Lv, Liang Hao, and Hui Lv, =93Resisting RST Watermarking Algorithm for Image Content Authentication,=94 2010 Second International Conference on Networks Security, Wireless Communications and Trusted Computing, Volume 2, Pages 361-364, April 2010.
[9] K A Navas, Ajay Matthews Cheriyan, M Lekshmi, S Archana Tampy, and M Sasikumar, =93DWT-DCT-SVD Based Watermarking,=94 3rd International Conference on Communication Systems Software and Middleware and Workshops, Pages 271- 274, January 2008.
image from Mark's project

Chad Griffith,
Isolating and Analyzing Rhythm Tracks For Use in Beat Detection Systems
Beat detection systems' core component is being able to generate reliable and acceptable rhythm tracks for use in beat analysis. However, there has not been a lot of research in the specific area of generating rhythm tracks, only the overall beat tracking system. This paper gives a short introduction on beat tracking methods, rhythm track generation, a short proposal for a middle man system, and short analysis of various rhythm tracks. Emphasis is directed towards more research in generating various quality rhythm tracks as well as manual and automatic analysis of these generated tracks.
image from Chad's project

Tariku Ahmed,
A Research on Frequencies out of Human Hearing Range Like Dog Whistle and Ultrasonic Pest
Record sound frequency made by dog whistle and Ultrasonic Pest Repellent, so that they become human audio-able. Based on this perhaps examine if it is possible to develop apps that could be used to irritate robbers not only that, but also if we could develop an app that would equipped our home security system, mobile phone, and car key chains to deter any intruders.

Sanghoon Lee,
Utilizing Morphological Processing for Moment Invariants
This paper focuses on the image preprocessing, especially morphological image processing for a variety of hand sign images. In order to provide the various hand signs, Kodaly hand signs are used in preprocessing step. The fathomable point in hand is seldom considered in the preprocessing. A gray-scale morphological technique is utilized in the preprocessing step to reduce the computational burden of feature extraction. This work has slightly lower error rates for extracting the feature vector than the processing on raw binary images.
image from Sanghoon's project

Guoliang Liu,
Image matching using area-based matching method(tested through MATLAB GUI)
Image matching method includes area-based matching methods or say template matching methods and feature-based methods. In this project, I realize five area-based matching methods relative to correlation-based and square deviation based methods. Through Matlab GUI, it is possible to test the accuracy and efficiency of each method actively and apparently. In the future work, matching in the video can be done but not in matlab since the image matching I implemented is not efficient. Double loops or more overlapped loops can be deadly in Matlab. In another aspect, Matlab GUI still have bugs and need to be be corrected and improved.
results slide from Guoliang's presentation

Zainab Haydari,
Signal analysis for differentiating healthy and pathological voices
In this paper I have used a method of voice signal analysis for differentiating healthy and unhealthy voices based on predictability measures. The met hod measures the predictive power (PP) that uses the relative entropy between signal and its prediction error to obtain the information about the predictability of the patient=92s voice production system, and is based on the wavelet decomposition. I have implemented this method and I have applied it to three groups of signals, i.e., healthy, nodule, and vocal cysts. The results show that the predictive power of pathologic groups are less than that of healthy group . This decrease is a result of increase of intricacy and irregularity of pathologic signals.
slide from Zainab's presentation

-Michael Weeks