Abstract: The green channel imbalance is a typical defect of Bayer image, which can be suppressed by correction of green balance algorithm, so as to reduce the maze noise of the color image after demosaicking, and improve the perception of image smooth area. With the development of image devices towards high resolution and large area array, the past green balance algorithm is difficult to meet the requirements of low delay and high pixel resolution. Therefore, a real-time green balance algorithm based on field programmable gate array (FPGA) was proposed, in which the smooth regions sensitive to the imbalance of green channels through local image gradients were extracted, thereby completing the green channel imbalance correction. While ensuring the resolution, the maze noises were removed, and the overall delay was less than the time of two lines of data transmission. The system design of front-end photosensitive camera, cache, image processing and final output image according to the features of FPGA hardware was completed, and the algorithm also could be compatible with other image processing functions. Experiments show that this algorithm can achieve microsecond-level delay while maintaining the resolution of pixel-level texture information of the image, and has excellent robustness for different levels of green channel imbalance defects and different scenarios.