Dissertation > Excellent graduate degree dissertation topics show

Image Registration Method Research Based on DM642 Processing Platform

Author: ZengZuo
Tutor: LiJianXun
School: Shanghai Jiaotong University
Course: Control Theory and Control Engineering
Keywords: Image registration Affine transformation Translation transform Affine morphological changes Scale space Harris corner
CLC: TP391.41
Type: Master's thesis
Year: 2010
Downloads: 273
Quote: 1
Read: Download Dissertation

Abstract


Image registration is an important field in image processing, it is looking for a change, different time, different sensors (imaging device), or under different conditions (weather, illumination, camera position and angle, etc.) to obtain two or more images match, the superposition of the process, it has been widely used in remote sensing data analysis, computer vision, medical image processing and other areas. Image registration methods are many, can be roughly divided into two categories, one is region-based methods, such as template matching method; the other is the feature point-based methods, such as Harris corner. In this paper, the algorithm of the above two types of methods focus on research. Registration algorithm for translational transform translational transform is a rigid transformation, the transformation matrix with two unknown quantities. This algorithm is based on the region, so the step of detecting a feature, directly on the image processing, the translational properties of the Fourier transform of the traditional method and a block matching method, but such methods are directly on the whole image processing efficiency is very low, making the next. This paper proposed an improved pan registration algorithm, which includes extraction algorithm and self-test template matching algorithm based on a priori information automatically prospects. This method first in the image to extract the foreground region of interest, and then selecting a gradation in the region characterized in having enough to distinguish the degree of templates and other images in the foreground area of ??the corresponding region for template matching. This method not only in the calculation amount than for the entire image has improved in accuracy at the same time is more reliable than the conventional block matching method. Registration algorithm for more complex scale transformation and affine transformation, this method is based on the type of feature, feature detection operator first find the feature points, Harris corner detection operator characteristics of a classic its illumination changes and rotation stability, but complex changes instability. To exist scale transformation of the image registration, the Harris corner extended to scale space, so that it has scale invariance. Order affine transformation of the image registration, the same point on the scale is modified until convergence, it has affine invariant affine form. These feature points, descriptors and matching algorithms using a common feature points to match point, and then use the matching points to calculate the geometric transformation matrix, thus completing the image registration of different perspectives. The experimental results show that the proposed improvements the Harris operator than SIFT operator can obtain more precise matching points and a smaller difference image matching. In recent years, with the continuous development of microelectronics technology, the rapid development of multimedia technology has been more widely used, especially video acquisition and processing technology has brought convenience to people's lives. DSP technology in recent years, the rapid development of a great help for the further development of image processing technology, TI Introduces TMS320DM642 digital media processor is dedicated DSP for video processing applications, it has a very rich hardware peripherals and outstanding processing performance image processing technology and DSP technology combine to build a DM642 DSP-based digital image processing platform with both visible and infrared imaging sensors. The experimental platform consists of multiple parts, including the DM642 DSP-based experimental evaluation board, in addition to visible light camera and an infrared camera to provide input display LCD TV. All of these hardware combined together to form a typical image processing hardware platform, based on the system to achieve the above two kinds of image processing algorithms. Monocular image stabilization target tracking system. Due to the characteristics of infrared imaging, infrared cameras and more used in military aircraft, satellites and other mobile platforms mounted on uncooled detector enables taking into account the sensitivity level sufficient to imaging a kilometer away, which is equipped with optical lenses are generally The telephoto lens is very sensitive to jitter, this particular application environment determines the characteristics of images collected image stabilization, so must its image acquisition. The system, first of all take advantage of the prospects for a common translation transform extraction, template matching method of infrared image image stabilization; then a series of pre-processing of the image of the target set by the user box, using the template matching method the target image position; Finally, experimental evaluation board serial PTZ PTZ control protocol control servo target area. 2. The binocular image barebones. Binocular vision, also known as stereo vision, it maximizes simulate human vision principle, the basic principle is the same scene is viewed from two viewpoints, and to obtain the image at different viewing angles, by the principle of triangulation to calculate the positional deviation between the image pixel access to the scene of the three-dimensional information. This process three-dimensional perception of the human visual process is similar. When the same object to be imaged by the two cameras at a certain angle, the imaging will certainly exist distortion, which requires the first two images of alignment to the two signal sources of the information fusion and then for further processing. This system uses two visible camera to shoot the same object at the same time a certain angle between the two cameras, the affine transformation between the image obtained. The system using improved Harris operator of two image registration under affine transformation, to prove the effectiveness of the operator.

Related Dissertations

  1. Research of Image Mosaic Technology,TP391.41
  2. Study on Image Registration and Image Matching of the Robot System for Substation Equipment Inspection,TP242.62
  3. Study on the Image Registeration Integrated with the LQ Model,R815
  4. Mean-Shift -based KLT and target tracking study,TP391.41
  5. Based on image analysis of self-service banking scene several intelligent security technology research,TP391.41
  6. The Next Generation Sequencing Data Processing,TP391.41
  7. Research of the Real-time Rock Deformation Monitoring System,TP274
  8. Image registration techniques based on the Scale Invariant Feature Transform (SIFT),TP391.41
  9. Research on Primary Corner Detection Algorithm,TP391.41
  10. Research on Image Matching Algorithm in Binocular Stereo Vision,TP391.41
  11. On Pixel-level Image Registration and Fusion Based on Wavelet,TP391.41
  12. Behavior detection in complex airport surveillance environment,TP277
  13. Research on Texture Mapping from Photograph,TP391.41
  14. Research and Application of Image Retrieval Technology Based on Feature Points,TP391.41
  15. Center Positioning of Target in Monocular Position and Attitude Measurement,TP391.41
  16. Surveillance video and playback video splicing technology research and implementation,TP391.41
  17. Study of Automatic Algorithm for Registration and Background Removal in Slice Images of a Mouse,TP391.41
  18. Research and Implementation of Non-rigid Medical Image Registration Method Based on Improved Optical Flow Model,TP391.41
  19. Rectangularized CSS Images and a Matching Algorithm,TP391.41
  20. Multispectral based on mutual information image registration algorithm,TP391.41
  21. Based on three-dimensional and four-dimensional affine transformation integer proposed chaotic image encryption system,O415.5

CLC: > Industrial Technology > Automation technology,computer technology > Computing technology,computer technology > Computer applications > Information processing (information processing) > Pattern Recognition and devices > Image recognition device
© 2012 www.DissertationTopic.Net  Mobile