/CMSC426_group

2021FallGroup Repository

Primary LanguageMATLABMIT LicenseMIT

CMSC426_group 2021 Fall Repository

Authors:

Yizhan Ao Yingqiao Gou

Project 1

Color Segmentation using GMM color segmentation, bayer filter, image acquisition, color-space

Estimation

Estimating p(Cl|x) directly is too difficult. Luckily, we have Bayes rule to rescue us! Bayes rule applied onto p(Cl|x) gives us the following: p(Cl|x)=p(x|Cl)p(Cl)∑li=1p(x|Ci)p(Ci) p(Cl|x) is the conditional probability of a color label given the color observation and is called the Posterior. p(x|Cl) is the conditional probability of color observation given the color label and is generally called the Likelihood.

Color Classification using a Gaussian Mixture Model (GMM)

In this case, one has to come up with a wierd looking fancy function to bound the color which is generally mathematically very difficult and computationally very expensive.

Project 2 Panorama Stitching

Stitching multiple images seemlessly to create a panorama

Project 3 Rotobrush

local classifiers, color confidence, shape confidence, local boundary deformation

Segmenting with Localized Classifiers

Initializing the Color Model and Color Model Confidence

To initialize the color model, we followed the process in Video Snap Cut and converted the input image to Lab color space.

Updating Local Windows

To estimate the large amounts of motion in the object, we used detectSURFFeatures, which is rotational invariant. We tried to force matching to focus on the foreground by removing the background (setting pixels to NaN) but this often resulted in the algorithm not finding enough matching points.

Estimating the Local Boundary Deformation

To track small bits of motion, the transformation alone was not enough. We used the optical flow to account for these small changes.

Updating Local Classifier

We first convert the image to Lab color space. For each window, we calculated the probability a pixel is foreground or background using pdf of the old GMMs and the current image.

Project 4 SfM or SLAM

Structure from Motion (SfM) or Simultaneous Localization and Mapping (SLAM)