Design And FPGA Implementation Of An Adaptive Threshold Edge Detection Algorithm

Edge detection in digital image processing is widely used in many areas, such as pattern recognition, computer vision, image segmentation, face recognition. Edge detection is an important step in these applications. The quality of edge detection has a great impact on the subsequent processing algorithm. After a long period of research in edge detection algorithm both in China and abroad, not only the noise resistant ability but also the quality of edge detection result has a great increasement. But a fixed threshold is normally used in many edge detection algorithms, which will affect the quality of edge detection. For this problem, in this thesis, we design an adaptive threshold edge detection system based on Sobel operator and realize it on a FPGA development board.The main contents and results include: 1. Consulting a lot of literatures about edge detection and it's FPGA realization and reviewing the research development of edge detection. Introducing some common edge detection algorithms and performing simulations using MATLAB. Through contrast and analysis, this paper proposes an adaptive threshold edge detection algorithm based on Sobel operator. 2. Designing fast median filtering algorithm and modified edge detection algorithm and realizing them on FPGA. Performing simulation to verily the feasibility and accuracy of these algorithms using modelsim. 3. Building a complete edge detection system including video images input and output on a FPGA development board based on an Altera's Cyclone IV serious FPGA EP4CE15F17C8. This system includes video images input module, SDRAM controller module and VGA display module. Using Verilog HDL to describe these modules. Performing functional simulation on these system using modelsim. 4. Verifying the complete edge detection system on the development broad. The experimental result shows that the proposed edge detection algorithm can not only extract edges in the video images correctly but also meet the requirement of real-time performance. For video images which has a big edge difference, this algorithm has a better edge detection result.