红外图像差分盒维分形算法的FPGA实现

Implementation of fractal algorithm of differential box dimension for IR image using FPGA

差分盒维分形算法作为一种红外图像分形维数提取算法广泛应用于红外图像目标检测跟踪算法中。针对其计算量大,单DSP无法单帧实时实现的缺陷,利用FPGA并行流水特点,采用模块化设计思想设计并实现了差分盒维分形算法,算法主要由四点极值计算模块和使能控制模块及对数计算模块组成。四点极值计算模块计算图像分割网格像素灰度极值,使能控制模块控制网格内像素灰度极值的计算开始时刻,而对数计算模块计算每一尺度下的分形维。仿真结果表明:采用FPGA实现差分盒维分形算法处理延时短,最高系统主频可达126 MHz,满足实时处理的要求。

Fractal algorithm of differential box dimension used for calculating fractal dimension of IR image,is widely applied in detecting the target in IR image processing.To resolve the problem that it can not be real-time implemented in one frame using one DSP for large calculation cost,an algorithm was designed and implemented with FPGA using modularization design method according to the characteristics of parallel and pipeline.The implementation consisted of three main modules： calculation module of minimum and maximum values for 4 points,enable control module and logarithm calculation module.The first module calculated the minimum and maximum of pixel gray value for segmented grids of IR image.The second module controlled the calculation start instant of pixel gray value.And the last module calculated the fractal dimension of each scale.The simulation results show that this algorithm can process IR image with shorter delay and the maximum main frequency of system is about 126 MHz.Therefore it can satisfy the real-time processing requirement.