This preview shows page 1. Sign up to view the full content.
Unformatted text preview: ns are often based on n-dimensional vectors of gray levels in n
spectral bands for each pixel or small pixel neighborhood. This segmentation approach,
widely used in remote sensing, results from a classification process which is applied to these
Based on local thresholding. Aim being to detect the presence of a region in a low resolution
image, and to give the region more precision in images of higher to full resolution. The
simplest method starts in the lowest resolution image (the highest pyramid level), applying
any of the segmentation methods discussed previously. The next step yields better
segmentation precision- pixels close to boundaries are re-segmented into either object or
background regions. This increase in precision is repeated for each pair of pyramid levels up
to the full resolution level at which the segmentation is obtained. A big advantage is the
significantly lower influence of image noise on the segmentation results, since segmentations
at lower resolution are based on smoothed data, in which noise is suppressed.
Another approach looks for a significant pixel in image data and segments an image into
regions of any appropriate size. The pyramid data structure is used again – either 2*2 or 4*4
averaging is applied to construct the pyramid. Faculty of Engineering Robotics Technology MECH 4041 9
B.Eng (Hons.) Mechatronics S. Venkannah Mechanical and Production Engineering Department Summary Thresholding Thresholding represents the simplest image segmentation process, and it is computationally
inexpensive and fast. A brightness constant called a threshold ids used to segment objects
Single thresholds can either be applied to the entire image (global threshold), or can vary in
image parts (local threshold). Only under very unusual circumstances can thresholding be
successful using a single threshold for the whole image.
Many modifications of thresholding exist: local thresholding, band thresholding, semithresholding, multi thresholding, etc.
Threshold detection methods ar...
View Full Document
This document was uploaded on 03/12/2014 for the course MECHANICAL 214 at University of Manchester.
- Spring '14