From the last couple of decades, image denoisification is one of the challenging areas in the image processing and computer vision domains that adds clarity to images by removing noise and makes them suitable for further processing. Images or videos captured by various visual sensors are degenerated during various circumstances such as sensor temperature, light intensity, target environment, erroneous electric instruments, and interference in the communication channels and required to be enhanced by various filtering techniques. Most of the real-world applications like student monitoring, vigilance system, traffic analysis, target detection, and classification use image smoothing to reduce artifacts present in the captured images and additionally demand its quicker execution. In this paper, we have explained different types of noise models normally degrade the quality of an image and the three most popular frequency domain smoothing filters (ideal, Gaussian, and Butterworth) are available to remove those noises. In addition to that, OpenCL kernels of those filters have been implemented for rapid denoisification and can be easily portable on wide verities of parallel devices. At last, performance evaluation has been carried out for both the CPU and GPU implementations to seek out their productiveness on those noise models in terms of time and precision. Various performance assessment metrics like entropy, root-mean-squared error, peak signal-to-noise ratio, and mean absolute error are applied to determine the rightness of the above filters. © 2021, The Editor(s) (if applicable) and The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd.