Seminar by Vinay Namboodiri
Exploring supportive cues for solving Computer Vision problems
Vinay Namboodiri
Bell Labs, Antwerp, Belgium
Date: Friday, November 2nd, 2012
Time: 5PM
Venue: CS101.Abstract:
This seminar describes the basis for solving computer vision problems by identifying and exploring appropriate supportive cues for solving computer vision problems. Specifically, three problems will be discussed.
The first problem is about estimating depth from defocused (blurred) images. In this topic, we will discuss a method proposed to estimate depth from a single defocused image. The key towards solving this lies in using the reverse heat equation to estimate depth in the image.
In the next topic, we will discuss the problem of super-resolving an image. Here, we will be discussing how learning of low level vision can be used to estimate the super-resolved image and how it can be adapted by using non-inform image patch exemplars.
In the last topic, we will explore the problem of classification of objects in images. We will discuss a method which proposes use of latent variables in terms of localization for classifying an object in an image.