Computer Vision News - December 2016

30 Computer Vision News Research Research Shapes From Pixels Every month, Computer Vision News reviews a research from our field. This month we have chosen to review Shapes From Pixels , a research paper studying the problem of recovering shape images with smooth boundaries from a set of samples. We are indebted to the authors ( Mitra Fatemi , Arash Amini , Loic Baboulaz , and Martin Vetterli ) for allowing us to use their images to illustrate this review. The full paper is here . Background: Sampling is a key procedure in digital signal acquisition. Continuous-domain signals are converted into a sequence of numbers to enable storage and processing of the data. With regards to digital image acquisition, the optical lens and the sensor array are responsible for the sampling process: the optical system determines the impulse response of the involved filter, which is called the Point Spread Function (PSF) ; the sensor array controls the number of samples and the sampling resolution. The purpose of this paper is to study the problem of recovering continuous-domain visual signals from their samples . Motivation: Recovering a continuous image will enable applications to have: (1) unlimited zooming capability; (2) the capacity to arbitrarily rotate the image; (3) improved feature extraction and object localization. Sampling system implemented by a digital camera; the effect of the optical lens and the sensor array on 2D visual signals can be modeled by filtering followed by sampling in space Recovering continuous-domain visual signals from their samples