04-09-2012, 02:56 PM
Image Processing : a DSP application
Image Processing.pdf (Size: 1 MB / Downloads: 129)
Introduction
An image is a two-dimensional signal whose intensity at any point is a function of two spatial
variables [4]. We note x and y the spatial coordinates and f (x,y) the image intensity values.
Examples: photographs, still video images, radar images, etc.
An image sequence (e.g. television) is three-dimensional signal since the image intensity at
any point is a function of the two spatial ones and the time.
Each picture element or pixel represents a certain physical quantity. For instance, a photograph
represents the luminances of various objects as seen by the camera. But an infrared image
collects the temperature profile values of a scene.
Image enhancement techniques emphasize specific image features to improve the visual perception
of an image.
Derivatives and smoothing
The previous methods are sensitive to the noise. In fact, edge filters are high pass-filters that
enhance spatial high frequencies such as edges but also noise. One solution is to decrease the
noise in the image by first applying a smoothing filter.
Using Difference of Gaussian (DoG) to approximate LoG
Another method for edge detection is to apply difference of gaussian to an image. Two gaussian
filters with different values of s are applied in parallel to the image. Then the difference of
the two smoothed instances ˜ f1 and ˜ f2 is computed. It can be shown that the DOG operator
approximates the LOG one. Human visual system uses a similar method to detect edges.
Applications
Sport Video Indexing
The emergence of multimedia technology coupled with the rapidly expanding data collection,
for private, industrial and public uses (e.g. self-made photos and videos repositories, Web resources,
etc.), have attracted significant research efforts in providing tools for effective retrieval
and management of the multimedia information. At first, retrieval techniques were mainly performed
using keyword index. This indexing technique, mainly done manually, has since shown
its limitation, both for its accuracy and for its applicability on huge database such as the worldwide
web where the daily growing data cannot be handled manually. In particular, motivated
by the commercial value of certain sports, retrieval and summarisation of sport events have received
increasing interest in recent years. Basic image processing techniques can help to extract
automatically relevant information from multimedia database.