Adaptive optics (AO) has been proved to be an invaluable technique to improve the imaging through turbulence for ground-based telescopes. By measuring the wavefront distortion in the input beam and consequently correcting in real time, modern large telescopes are able to approach their diffraction limits. One of the difficulties limiting the common adoption of AO for small size telescopes arises from the requirement of wavefront sensing at a high speed to match the temporal bandwidth of the turbulence and its associated complexity in computation and cost. The holographic modal wavefront sensor (HMWFS) promises a simple, high speed and low cost wavefront sensing scheme, where the modal contents of the input aberrations can be measured directly by calculation of intensity ratios between the pairs of spots reconstructed from the hologram. It has attracted much attention, due to these benefits, from the research communities in the past 10 to 15 years, but mainly discussed under numerical simulation and limited laboratory test under coherent laser sources. To explore its usability in AO systems for astronomy, we demonstrated the necessary improvements from an engineering perspective. An improved detection method was proposed to improve the response from the HMWFS, with dedicated hardware implementation to guarantee the detection speed. Common holography techniques are applied to the construction of HMWFS to better meet the foreseen challenges of hologram reconstruction under wide spectrum and low photon star light. Field-programmable gate array (FPGA) devices have been regarded as an ideal platform to implement the heavy computation tasks in the wavefront sensing and reconstruction processed due to its powerful parallel processing capabilities. In this thesis, we developed a stream-based centre of gravity (SCoG) algorithm to estimate the centroid of spots and track their movements automatically. The algorithm was mainly evaluated in a Shack-Hartmann wavefront sensor due to its superiority in terms of reducing the noise errors and expanding linearity and dynamic range. However, it is general enough to be extended for other applications such as supporting the HMWFS or measuring the tip/tilt by tracking the focal spot. An FPGA implementation of the SCoG algorithm has been tested both in laboratory experiments and on-sky observation to validate its usability.