PRESENTATION REU IN COMPUTER VISION 2014 AMARI LEWIS CRCV UNIVERSITY OF CENTRAL FLORIDA
IMPLEMENTING DIFFERENT WAYS TO IMPROVE PICTURES… Original The top image combines the different channels and uses convolution F *h= Σ Σ f(k,l)h(-k,-l) F= image H=kernel
COMBINE CHANNELS
GAUSSIAN Type of smoothing, a weighted average of the surrounding pixels using this formula: The sigma value determines the amount of ‘blurr’ the image will display. Gaussian smoothing Original
‘LAPLACIAN’ Finds the 2 nd Derivative of Gaussian
HISTOGRAM – USED TO REPRESENT EACH COLOR IN THE IMAGE OBSERVE BELOW
EDGE DETECTION- Roberts Roberts: finds edges using the Roberts approximation to the derivative. It returns edges at those points where the gradient of I is maximum. Canny Uses two thresholds to determine between weak and strong edges Canny Roberts
EDGE DETECTION WITH THRESHOLD Sobel X: [1 0 -1, , ] Y: [1 2 1, 0 0 0, ] Calculates: √(d/x)²+(d/dy)²
PYRAMIDS
ADABOOST – FACE DETECTION Boosting defines a classifier using an additive model F(x) = ∂1f1(x) +∂2f2(x)+∂3f3(x)…. F:strong classifier X- feature vectors Sigma= weight f – weak classifiers
TRIAL 2
SVM SVM (Support Vector Machine) classifier is able to test trained data to analyze and divide results. (object ore non—object) This is an example of linear classification Linearsvm calculates : f(x) = w^Tx+b where w is the normal line or weight vector and b is the bias
RESIZING MULTIPLE IMAGES THROUGH FOR LOOPS..
LUCAS KANADE (LEAST OF SQUARES) Optical flow equation- Considers a 3x3 window
Lucas Kanade
OPTICAL FLOW
LUCAS KANADE WITH PYRAMIDS
CLUSTERING, BAG OF FEATURES
THE PROJECT I’M INTERESTED IN WORKING ON THE APPLICATIONS OF LIGHT FIELDS IN COMPUTER VISION AIDEAN SHARGHI
THANK YOU !! I APPRECIATE THE OPPORTUNITY ONCE AGAIN AND I AM LEARNING A LOT FROM THIS EXPERIENCE THANKS, OLIVER NINA DR. LOBO DR. SHAH