Sample+5

=Project Title: Enhance Images while Maintaing Clarity and Minimizing Distortion.= =External Advisor: Dr. Pang Sze Kim= =Mentor: Mr Joseph Tan Choo Kee= =Done by: Low Rui Hao (4S2), Caleb Foo (4S1)=

Transforming images to fit different purpose have been widely used, whether it is for a design, orfor large billboard prints. Taking images with mobile phone cameras have always resulted inpixelated and blurred images which is uncomparable, in terms of clarity, to images taken withcompact cameras or Digital SLR cameras. Transforming the images by scaling it down also causes distortion. The aim of the project is to create a program whereby images are enhance, as well as minimizingdistortion and maintaing clarity. The purpose of the mentioned project is to enhance imageresolution as well as to minimize distortion while transforming for different uses. A specific part of the project is minimizing distortion, which require one concept, namely **Anti-Aliasing**. Anti-aliasing is the technique of minimizing distortion when representing a highresolution image at a lower resolution. It is utilised in many applications including digitalphotography and computer graphics. In such technique, supersampling with the Quincunxes pattern is able to combine five sample points, in the shape of the a quincunx, to produce each displayed pixels. Quincunx anti-aliasing could not only be used in images, but also in 3D images, with objects producing the “staircase effect” of jagged edges. As such, anti-aliasing is one of the key concept in minimizing image distortion. Besides using supersampling in Anti-aliasing, sparse sampling may also be applied. In thesesampling, the prior that the signal is sparse in a basis or in a parametric space is put to contributionand perfect reconstrution is possible based on a set of suitable measurements. The idea behindmulti-frame image ** super-resolution **is to combine several blurred low-resolution images or videoframes to produce a single detailed high resolution image. Super resolution techniques thus allow to overcome the hardware limitation of the acquisition devices and to obtain images that wouldotherwise require much more expensive hardware (Dragotti, 2009). A recent testing of this methodwas conducted with 60 images aquired from a Nikon D70 DSLR, each 60*60 pixels. The resultedproduct was a super resolved image of 600*600 pixels which is clearer and higher in resolution. In resizing images and maintaing clarity, a concept would be digital image interpolation.Interpolation works by using known data to estimate values at unknown points. Besides resizingimages, rotating images do cause image degradation, thus it is advisable not to rotate more thanonce. There are many algorithms in Image Interpolation, such as non-adaptive and adaptivealgorithms. One of the key algorithm is the bicubic algorithm, a non-adaptive one, which producessharper images. ** Bicubic algorithm ** inlcudes bicubic smoother and bicubic sharper. Bicubic smoother algorithm would be more effective when use in increasing size of image, while Bicubic Sharper is more effective for reducing the size. Image restoration, is one of the topic many research upon. In a study conducted by L. Mancera //et al in 2009, they created a new image restoration method via mixture modelling of an overcomplete //Linear Representation. Their experimental results shown better image produced, whereby theyused the Peak Signal-to-Noise Ratio (PSNR) to measure restoration quality numerically. Theirmethod, however, includes more fine tuning and future works including estimation of the optimalalgorithmic parameters and incorporation of local adaptivity for better texture recovery. In image enhancement, both anti-aliasing and image interpolation is able to work together to bettercreate a sharper as well as smooth-edge image. As such, a program with both algorithms as well as to improve it would be the product, to further enhance images. Proposed Timeline is as follow:
 * Time || Activity ||
 * T1 W5 || Now ||
 * T1 W6-8 || Meet Up with External Mentor on progress ||
 * T1 W9-March Hols || Understanding algorithms and translating into C++ ||
 * Buffer || Buffer ||
 * T2 W1 || Meeting up with Mentor and External Mentor on progress, advise ||
 * T2 W2-4 || Testing out program with image, setting parameters ||

=
============= __Methodology of Project/ Plan of Action__ On meeting up with our external advisor, Dr. Pang on 26 February 2010, Friday, we were advised to have a clear layout of our plan of action for the above stated project. The main focus of the project would be the learning of how to enhance the graphics and images with certain methods and algorithms, and too much focus placed on programming is not advisable. As such, the following programs available online are advised by Dr. Pang: 1. Octave [link: __[] ]__ Octave is a computer program for performing numerical computations, and it is mostly compatible with MATLAB. It has matirces as fundamental data type and have built-in mathfunctions and an extensive function library. 2. MATLAB [link: __[] ]__ 3. OpenCV (Open Source Computer Vision) [link: __[] ]__ OpenCV application extends to image, object, gesture recogniton, motion tracking, and has a statistical machine learning library that contains many algorithms such as // k-nearest neighbour algorithm //. It includes C++, C and Python, and hence easier to picked up These programs could be downloaded on our own laptops, or could most probably be downloaded in the VR lab, for usage during project meeting sessions. __[|**Methodology**]__ This project is expected to be near completion by the start of Term 3, preferably the first week of July. Members of the project however have the whole of June Holidays, as well as allocated time slots in the afternoon during school curriculum to complete. As such, the project is going to be broken down into several __ modules __ : Module 1 is the basic of the project. Module 2 to 4 will be researched upon at each stage, to determine the various algorithms used by many people, and then to determine the best algorithm to used based on certain factors, such as outcome or an easier algorithm to be understood by many. Module 5, however, could be a // further work // and should be done only if time allows. These modules should be done in order so as to ensure that work are done, // i.e. Module 1 to 3 //, even if time is short.
 * Programs / Software**
 * 1) Opening an Image in a Program
 * 2) Anti-Aliasing
 * 3) Bicubic Interpolation
 * 4) Super resolution by Sparse Sampling
 * 5) Seam Carving for Image Resizing

=
=====================

References (n.d.). Anti-aliasing. Retrieved from http://en.wikipedia.org/wiki/Anti-aliasing

Thomas, Pabst (2001, February 27). Anti-Aliasing-Removing The 'Jaggies '. Tom's Hardware Guide, Retrieved from http://www.tomshardware.com/reviews/high, 294-23.html Sean, McHugh(n.d.) .Understand Digital Image Interpolation. Retrieved from http://www.cambridgeincolour.com/tutorials/image-interpolation.htm(n.d.).

Resizing- Bicubic or Bicubic Smoother. Retrieved from http://www.photoanswers.co.uk/Community-Landing/Forum-Landing/Forum-Categories/Topic/?topic-id=17062&message-id=83432&__ia=message83432#message83432# message 83432

Dragotti, P.L. (2009). Sparse sampling and its application to image super-resolution.

Mancera, L., //et al// (2009). Image restoration by mixture modelling of an overcomplete linear representation.

Image Progress: 1. Phase 2: Draft 1

2. Phase 2 note: Have an image dimension less than 1000px and rename it "testimage2"

Files and Documents 1. 2. 3. 4. 5. 6. C++ code for opening a BMP image 7. C++ code for opening a GIF and BMP image

Journals 1. 2. 3. 4. 5. Till date, Journal 1, 4 and 5 is the most relevant to our Project.