Font Size: a A A

System Modeling Of A Scale-invariant Image Feature Point Detector In Systemc

Posted on:2010-10-22Degree:MasterType:Thesis
Country:ChinaCandidate:Y F SunFull Text:PDF
GTID:2178360332457888Subject:Microelectronics and Solid State Electronics
Abstract/Summary:PDF Full Text Request
Image feature detection is a cocept from computer vision and image processing. As a first step for image processing, image feature detection is widely used in object recognizaton, mobile robot location, image mosaic and image registration. Lowe has proposed a scale-invariant feature transform algorithm with abbreviation SIFT based of previos studies in image featrue detection. This algorithm generates rotation and scale invariant feature descriptors, which are also partically invariant to brightness changes, contrast changes and affine transform.The main purpose of this dissertation is supplying an image feature point detection method implementation for the medical image regristration problem which is involved in project of three-dimentional reconstruction of the CT images. As SIFT algorithm is robust in computational cost and corret repeatability, it is widely applied. This dissertation studies the system level modeling of a feature point detection system which applies SIFT algorithm in SystemC. Firtly, SIFT algorithm is studyed and simulate with matlab. Then the feature point detecting system is divided into models according to the dataflow in SIFT algorithm. In the end, the functional correction of the SystemC model is verified with a gray level image as the input test source.The SystemC models are highly configurable, thus are helpful to system architechture optimization and design space exploration. SystemC does not only support system level simulation, but also support register-transfer level modeling. In this dissertation, a gaussian convolution model is refined from system level to behavioral level. Based on the studies of this dissertation, the feature point detecton system can be implemented with FPGA by refinement of each model. The excutable file of the feature point detection model generated by SystemC could also be integrated into the software enviroment for medical image registration.
Keywords/Search Tags:feature detection, SIFT, SystemC, sytem level modeling, scale space
PDF Full Text Request
Related items