Research And Implementation Of Image Edge Detection Algorithm Based On FPGA

With the maturity of multimedia applications and the rapid development of computer vision,digital image processing technology has been applied to almost all fields.In fact,most of the algorithms for image processing are based on software platform simulation,so there is no guarantee that it can perform complete function mapping implementation on the hardware platform.This paper mainly studies the image edge detection algorithm based on FPGA.Starting from the core part of image processing(two-dimensional convolution),it will realize its function on FPGA,and then analyze various image processing effects from the perspective of convolution kernel.Based on the mapping improvement of the Cannyr model for image edge detection,the Sobel model is improved which is a part of the Canny model and this Sobel model will be verified.The research done in this paper is as follows:1.The image processing scheme based on FPGA realizes two-dimensional convolution,based on 3×3 convolution kernel,performs three-dimensional multiplication weighted summation through three buffers on FPGA,and finally accumulates three weights and realizes two-dimensional convolution processing.Taking 2D convolution processing as the core,processing RGB images,improving the reading mode of RGB images,and comparing the processing effects of RGB images on FPGA and MATLAB under various thresholds The effectiveness of two-dimensional convolution image processing based on FPGA.2.The traditional Sobel model is improved by Cordic technology.The main function of the Sobel model is to realize the desired edge detection by calculating the gradient information of the image.In this paper,we use the vector pattern in the Cordic calculation model to transform the Cartesian rectangular coordinates into the polar coordinate system,and calculate the modulus and gradient information through the iterative after rotation to realize the function of the Sobel model,thus avoiding the design of FPGA.The IP core is called to complete the operation of finding the root number and the inverse tangent,which greatly saves the on-chip resource occupancy rate.3.Call the improved Sobel model in the Canny model.Firstly,the Canny model uses Gaussian filtering to complete the smoothing of the image,call the improved Sobel model and pass the processed direction information through the lookup table.Secondly,do non-maximum suppression processing on the modulo value in the calculation result,here the equation transformation elimination operation is combined with the previously stored lookup table to achieve the consistency of the calculation,improve the calculation efficiency,and delay the final processing result.Finally,the threshold processing completes the image edge detection process of the entire Canny operator.