Image Processing Techniques

B. Voc. in Medical Imaging Technology

VERSION 2
 Introduction
In general, image processing techniques are based on three types of
operation: point operations (point processes), local operations (area
processes), and global operations (frame processes).
The image processing algorithms on which these operations are based
alter the pixel intensity values.
The exception is the geometric processing algorithm, which changes the
position of the pixel.

VERSION 2
 Point Operations
Point operations are perhaps the least
complicated and most frequently used image
processing technique.
The value of the input image pixel is mapped
on the corresponding output image pixel.
The algorithms for point operations enable
the input image matrix to be scanned pixel by
pixel, until the entire image is transformed.

VERSION 2
 Point Operations (Contd.)
The most commonly used point processing
technique is called gray-level mapping.
This is also referred to as "contrast
enhancement," "contrast stretching,"
"histogram modification," 'histogram
stretching," or '"windowing,".
Gray-level mapping uses a lookup table (LUT),
which plots the output and input gray levels
against each other.
.
VERSION 2
 Point Operations (Contd.)
LUTs can be implemented with hardware or software for gray-level
transformation.
In the transformation process, the gray level mapping changes the brightness of
the image and results in the enhancement of the display image.
Gray-level mapping results in a modification of the histogram of the pixel
values. A histogram is a graph of the pixels in all or part of the image, plotted as
a function of the gray level.
A histogram can be created as follows:
◦ Observe the image matrix and create a table of the number of pixels with a specific
intensity value.
◦ Plot a graph of the number of pixels versus the gray levels (intensity or density values).
VERSION 2
 Point Operations (Contd.)
Histograms indicate the overall brightness and contrast of an image.
If the histogram is modified or changed, the brightness and contrast of the
image can be altered, a technique referred to as histogram modification, or
histogram stretching.
This is also an example of a point operation in digital image processing, if
the histogram is wide, the resulting image has high contrast.
If the histogram is narrow, the resulting image has low contrast.

VERSION 2
Point Operations (Contd.)

VERSION 2
 Point Operations (Contd.)
On the other hand, if the histogram values are closer to the lower end of
the range of values, the image appears dark, as opposed to a bright image,
in which the values are weighted toward the higher end of the range of
values.

VERSION 2
 Local Operations
A local operation is an image processing
operation in which the output image pixel
value is determined from a small area of pixels
around the corresponding input pixel.
These operations are also referred to as area
processes or group processes because a group
of pixels is used in the transformation
calculation.

VERSION 2
 Local Operations (Contd.)
 Spatial frequency filtering is an example of a local operation that concerns
brightness information in an Image.
If the brightness of an image changes rapidly with distance in the horizontal; or
vertical direction, the image is said to have high spatial frequency.
When the brightness changes slowly or at a constant rate, the image is said to
have low spatial frequency.
Spatial frequency filtering can alter images in several ways such as image
sharpening, image smoothing, image blurring, noise reduction, and feature
extraction (edge enhancement and detection).
There are two places to perform spatial frequency filtering:
◦ In the frequency domain which considers the Fourier transform, or
◦ In the spatial domain, which uses the VERSION
pixel2values (gray levels) themselves.
 Local Operations (Contd.)
Convolution, a general purpose algorithm, is a technique of filtering in the space domain.
The value of the output pixel depends on a group of pixels in the input image that
surround the input pixel of interest: in this case, pixel P5. The new value for P5 in the
output image is calculated by obtaining its weighted average and that of its surrounding
pixels.
The average is computed using a group of pixels called a convolution kernel, in which
each pixel in the kernel is a weighting factor, or convolution coefficient. In general, the
size of the kernel is a 3 x 3 matrix. Depending on the type of processing, different types of
convolution kernels can be used, in which case the weighting factors are different.
During convolution, the convolution kernel moves across the image, pixel by pixel. Each
pixel in the input image, its surrounding neighbours, and the kernels are used to compute
the value of the corresponding output pixel—that is, each pixel is multiplied by its
respective weighting factor and then summed.
VERSION 2
 Local Operations (Contd.)
The resulting number is the value of the centre output pixel. This process is
applied to all pixels in the input image; each calculation requires nine
multiplications and nine additions. This can be time consuming, but special
hardware can speed up this process.

VERSION 2
 Global Operations
In global operations, the entire input image
is used to compute the value of the pixel in
the output image.
A common global operation is Fourier
domain Processing or the Fourier transform,
which uses filtering in the frequency domain
rather than the space domain.
Fourier domain image processing
techniques can provide edge enhancement,
image sharpening, and image restoration.

VERSION 2
 Geometric Operations
Geometric operations are intended to modify the spatial position or
orientation of the pixels in an image.
These algorithms change the position rather than the intensity of the
pixels, which is also a characteristics of point, local, and global operations.
Geometric operations can result in the scaling and sizing of images and
image rotation translation.

VERSION 2
 Image Processing Hardware
A basic image reprocessing system consists of several interconnected
components. The major components are the analog-to-digital converter,
image storage, image display, image processor, host computer and digital-
to-analog converter.

VERSION 2
 Image Processing Hardware (Contd.)
Digitizer: The analog signal is converted into digital form by the digitizer, or ADC.
Image memory: The digitized image is held in storage for further processing. Several
components are connected to the image store and provide input and output. The size of this
memory depends on the image.
Digital-to-analog converter (DAC): The digital image held in the memory can be displayed
on a television monitor. However, because monitors with analog signals, it is necessary to
convert the digital data to analog signals with a DAC.
Internal image processor: This processor is responsible for high-speed processing of the
input digital data.
Host computer: In digital Image processing, the host computer is a primary component
capable of performing several functions.
The host computer plays a significant role in applications that involve the transmission of
images to another fixation, such as medical imaging.
VERSION 2
 CT as a Digital Image Processing System
A number of imaging modalities in radiology use image processing
techniques, including digital radiography and fluoroscopy, nuclear
medicine, magnetic resonance imaging, ultrasound, and computed
tomography.
The future of digital imaging is promising in that a wide variety of
applications have received increasing attention, such as 3D Imaging.

VERSION 2
 CT as a Digital Image Processing System (Contd.)
The steps in image digitisation which can be applied to the CT process are as
follows:
◦ First, divide the picture into pixels. In CT, the slice of the patient is divided into small regions
called voxels (volume element) because the dimension of depth (slice thickness) is added to
the pixel. The patient is scanned as the x-ray rube moves around the patient.
◦ Next, sample the pixels. In CT, the voxels are sampled when x-rays pass through them. This
measurement is performed by detectors. The signal from the detector is in analog form and
must be converted into digital form before it can be sent to the computer for processing.
◦ The Final step is quantization. In CT, the analog signal is also quantized and changed into a
digital array for input into the computer. The digital data resulting from quantization are
processed by the computer through a series of operations or techniques to modify the
input image. In CT, the digital data are also subject to several processing algorithms so that
the output image can be displayed in a form suitable for human observation.

VERSION 2
 References
Computed Tomography by Euclid Seeram, Saunders; 3rd edition
(November 13, 2008)

VERSION 2