Abstract: Mean filtering algorithms are widely used for image processing especially with image smoothing or denoising systems but its hardware implementation has many restrictions. Filtering Process needs image segmentation that was done by pipelined windowing technique to scan the processed image horizontally and vertically and the size of this window determines the length of mean filter. All pixels within the specified window are drawn simultaneously and processed together in each cycle. This process produces repeated calculations that leads to depleting more hardware which can be reduced highly if an efficient implementation method is used. Another important problem with traditional implementation method of this filter is the need of more I/O pins, since all pixel values of the processed window must be read at the same cycle. These I/O pins act as a big limitations that restrict the huge data manipulation systems when implemented with any high speed hardware platform. In this paper, two novel techniques were proposed to design and implement mean filtering algorithms, that ensure simplicity of the required hardware and efficiently improve processing time as well as the needed input pins are minimized highly by reading only the new pixels of the processed window, while keeping the remaining big part as it is without any change, since it is already prepared in previous cycle. The proposed techniques depend mainly on removing the repeated computations and cancellation of division operations which result in improving the processing efficiency and highly reduce the execution time to match real time requirements. Pipelining technique is also adopted here, with the second proposed technique, to activate parallel processing scheme. So, systolic architecture were used for further reduction in computation operations and to improve overall system execution time. The obtained results proved good enhancements in processing time especially for huge image size. The efficiency of implemented design was significantly improved for large filter length, that makes the proposed techniques very useful when dealing with huge data systems. The effect of Gaussian noise was tested for gray and colored images to compare the degradation of both of them with this type of noise.