|
Philips Applied Technologies has been a major contributor
to the development of the advanced image-processing
techniques currently used in Philips high-end TVs –
including development and implementation of Philips’ award
winning Pixel Plus, Digital Natural Motion and Luminance
Transient Improvement picture-enhancement technologies.
Significant reductions in image processing time
Recently expertise has been built up in image processing
on the PC platform, in particular in exploiting the
processing capability of the PC’s graphics processing unit
(GPU) for image processing. Designed specifically for 3D
video rendering, today's powerful GPUs are generally
under-utilized in most PC applications. Transferring at
least some of the image processing tasks to the GPU can
therefore lead to significant reductions in image processing
time – often sufficient to allow the image processing to be
performed in real-time. It also frees up capacity on the CPU
for other tasks.
 One
area set to become a major beneficiary of these developments
is medical imaging which already has huge image processing
requirements – currently met by high-performance Pentium PCs
supported by advanced (and expensive) customized accelerator
cards. Moreover, these requirements are continually
increasing to meet the demands for better pictures, higher
frame rates and more advanced signal processing to extract
the information needed for diagnosis. Anticipating these
developments, Philips Applied Technologies is now making its
image processing expertise available for medical imaging
applications. Simulations using a reference algorithm ported
to the PC GPU have already been run. This has shown that
significant processing improvements are to be expected.
Advanced segmentation algorithms
 Moreover,
over the last years, we have built up a lot of
machine-vision competences for a variety of applications –
including presence detection, object recognition and
geometric measurements. The advanced segmentation algorithms
essential for such applications can also be easily reused in
medical applications, where many types of object
segmentation are required. For example, the size, volume or
boundaries of objects often need to be determined for
quantitative measurements, and since manual segmentation is
a tedious and time consuming process, automated or assisted
segmentation can significantly improve the workflow and
reduce costs.
To support this, Philips Applied Technologies is also
making its segmentation expertise available for medical
imaging applications and recently conducted a feasibility
study in which the segmentation framework on ultrasound
tumor and vessel images was tested. Preliminary results are
very promising.
|
news loading