Dissertation > Excellent graduate degree dissertation topics show
Research on Local Optical Flow Estimation Methods
Author: YuDanDan
Tutor: QuZuoShen
School: Harbin Institute of Technology
Course: Control Science and Engineering
Keywords: Optical flow calculation Horn-Schunck algorithm Lucas-Kanade algorithm No texture Discontinuous movement
CLC: TP391.41
Type: Master's thesis
Year: 2011
Downloads: 61
Quote: 0
Read: Download Dissertation
Abstract
|
Optical flow estimation has always played a very important role in the motion image analysis . Already most of the optical flow estimation algorithm can not complete local area pixel motion estimation . Based on the image texture area (untext) and the movement is not continuous at the regional (disc) study to design a new algorithm to reduce the calculation error of the original algorithm in these areas . This paper introduces the matching algorithm and other algorithms , and design a new method of calculation of the neighborhood better motion estimation results , given in order to improve the calculation accuracy . The main content of the paper are as follows : First, we study the basic methods of optical flow estimation , to grasp the basic idea of several optical flow algorithm , and simulation . Horn-Schunck algorithm regularization method to strike a light flow , but calculation error in the edges of the image area . Lucas-Kanade algorithm using local ideas , the introduction of least-squares algorithm to strike a light flow , but poor in the poor area of image texture algorithm accuracy . Secondly , the paper texture area and the edge of the area proposed to improve the program . Through research and analysis , the majority of the optical flow algorithm on most of the pixels on the image can not be correctly estimated . Such pixels are mostly distributed in the non - texture region and a motion discontinuous region . Therefore introduced in the non - texture region on the pyramid algorithm and block matching algorithm , assuming that each of the blocks to make the translational movement , in order to non-textured region on the optical flow of pixels . For movement is not continuous on the pixel in the region , since the region on the pixel motion movement in most cases inconsistent, even the existence of the pixel of the multi-movement , and therefore in such a special area , the use of the edge region of the image edge algorithm determined , and With this paper, two neighborhood to determine the method new neighborhood of computing and the center pixel motion consistent point , the use of a new neighborhood to strike the pixel optical flow . Finally, the improved algorithm simulation experiments . Experiments show that these improvements, can solve the the majority pixel optical flow to solve the problem , to get more accurate results of the optical flow , by comparison with the original algorithm , the improved algorithm on absolute error and angle error indicators , both the improved.
|
Related Dissertations
- Video Coding Research Based on Texture Characteristics,TP391.41
- The Research of the Text Extraction Method Based on Spectral Cut,TP391.41
- Research on Recognition of Fabric Weave Pattern Based on Space-Frequency Domain,TS101.923
- Studies on Changes of Quality Characters and Prediction Model of Postharvest Tomato Fruit,S641.2
- The Research on Texture Synthesis Technology from Cloud Theory & Been Evolution Genetic Algorithm,TP391.41
- Information Extraction of Marine Oil Spill with Collaborative Images of ASAR and MODIS,X87
- Cotton Different Texture Combination in Women’s Fashion Design Application Research,TS941.2
- Study on Microstructure and Texture during Hot Rolling and Normalizing of Twin-roll Cast Silicon Steel Strip,TG335.11
- Inspection and Recognition on a Class of Tool Processing Trace,TP391.41
- The Texture Research on Low Grade Non-oriented Silicon Steel,TG142.1
- Rolled Mg-Zn-Ce-Zr (Mn) alloy microstructure and mechanical properties of,TG339
- Mammary gland lumps Texture feature extraction and auxiliary diagnostic methods research,TP391.41
- Research on Application of Fractional Differential to Digital Image Processing and Support Vector Machine to Face Recognition,TP391.41
- Habitation Extraction and Edge Optimization,P237
- Effect of Rare-earth Elements on Dynamic Recrystallization and Texture of AZ31 Wrought Magnesium Alloy during Hot-rolled,TG339
- Research on Microstructure Evolution in Pre-heat Treatment and Cold Ring Rolling,TG142.15
- Reach on Pressure Processing for Producting High Pure Aluminum and Ce and Gd Target Materials,TG306
- Ecological perspective sewer network landscape texture of small towns,TU986
- Elevator guide shoe sliding friction system research and application,TU857
- Study of Image Feature Extraction and Texture Classification Algorithm,TP391.41
- Texture Based Fractal Imagery Modeling, Rendering and Morphing Algorithm,TP391.41
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
|