|
Researches on edge detection, contour extraction and image
segmentation have been an important area in image processing.
In recent years, many methods has been established including
space-filters, region growing, graphic search and active
contour model (snake model), etc. Each method has its
characteristics for suitable images. Proper method should be
chosen depends on different needs for different imaging
modality. Because ultrasound images(UI) are complex, blurring
and noisy, traditional methods are not adaptive to analyze UI.
The purpose of our research is to develop a novel method to
actively extract the contour in UI. It is convenient for
doctors to acquire some diagnostic parameters(for example, area
for clinical purposes. ). In the part, many studies using the
snake model applied to different kinds of images including
photography, medical images, dynamic images etc. But for UI,
there is not a good systematic approach to extract the contour.
after our investigation, we adopt the snake model as a basis,
combining adaptive-windows-filter, adaptive adjusting
weightings and integral amplifier to detect edges. The adaptive-
windows-filters can suppress the noise and preserve the edge
signal, which is superior than the traditional smoothing
filters. By automatically adjusting the snake weightings make
the snake model adaptive. The integral amplifier can amplify
the edge signal, and detect the ramp edge. This novel snake
model we develop performs better than traditional snake.
Applying the new snake to phantom simulation and clinical UI,
we all got satisfactory performance. The future works are four
folds : 1. Modified the function of filter or study another
better method. 2. Design better cost function making the snake
powerful. 3.Optimize the performance. 4.Apply snake to
communication and 3-D images reconstruction.
|