The invention discloses a method for hardware acceleration of an image processing function in an OpenCV library. The method is performed according to the following steps: A, defining a function pointer type in a file where a function requiring hardware acceleration is located; B, defining a static global variable of the function pointer type; C, defining an assignment function to assign a value to the static global variable; D, modifying the function requiring hardware acceleration and invoking a function pointer of the static global variable; E, recompiling the OpenCV library to generate a dynamic link library; F, defining a hardware acceleration function in OpenCV application codes, providing the completed functions the same as functions of the original part of the function requiring hardware acceleration in the library; G, invoking the assignment function and transmitting the hardware acceleration function to the assignment function as a parameter; and H, changing the hardware acceleration function defined in step F to be implemented by FPGA hardware logic. According to the invention, the running speed is effectively increased.