Storage Size Reduction for the Integral Image
Year: 2007
Type: article
Abstract: The integral image is an image containing accumulated sums of pixel values taken from an input image. It is an important concept for the computationally efficient implementation of so-called box filters, where the sums of the values of pixel in a rectangular image region are calculated. An important example of an algorithm using the integral image, treated in this report, is the well-known Viola and Jones object detector. Though computationally efficient for its low number of operations, unaddressed issues with the integral image are its large storage size and word length. This report deals with methods to reduce both considerably. As a result the integral image becomes an even more attractive concept for applications having memory and power constraints, like portable multi-media devices. One algorithm is taken as example for an experiment, namely a Viola and Jones face detector. It is shown that with the proposed word length reduction technique itis possible to apply Viola and Jones face detection on VGA image using only a 16-bit media processing architecture.
Authors: H. J. W. Belt, H. J. W. Belt, Frits De Bruijn
Language: English
Subfield: Artificial Intelligence
Field: Computer Science
Domain: Physical Sciences
Open Access status: green
OpenAlex ID: W1559647261