High-frequency (HF) ultrasound imaging, capable of achieving superior spatial resolution in real-time, has been shown to be useful for imaging and visualizing blood flow in ophthalmology, dermatology, small animal and intravascular imaging. The utilization of HF linear arrays along with beamformer for beam focusing can alleviate the limitations of the systems with single element transducers, such as single focal point, image degradation caused by mechanical and limited frame rate. This dissertation presents an investigation into the development of a high-frequency linear array ultrasound imaging system using analog beamformer. A HF high frame rate duplex system with 30 MHz 64-element linear arrays capable of fast B-mode image and pulsed-wave Doppler measurement was first implemented using high speed electronics to provide real-time duplex visualization for small animals cardiovascular activities. Subsequently a 64-channel analog beamformer was next developed in order to achieve better lateral resolution. In this design new passive delay line components can provide a total delay of 157 ns with 1ns resolution to support a 30 MHz 256-element linear array with 50 μm pitch. The transient effect caused by multiplexer switching was alleviated by using low charge injection multiplexers, differential circuits and time domain offsetting. An imaging system based on this beamformer was built by integrating it with the transmit beamformer, pulser/receiver, and a PC back-end controlling and processing unit implemented using a powerful and low-cost graphics processing unit (GPU). Both phantom and in-vitro experiments demonstrated that the system was able to achieve better resolution and clearly reveal the detail structures of the biological tissue samples such as excised bovine eye. High-frequency color flow imaging (CFI) using linear array was also investigated. In order address the challenge in high-frequency CFI, an adaptive clutter filter was proposed and the experimental results indicated that the adaptive filter could more effectively suppress the clutter signal under the high-frequency condition. The linear array system was then expanded to support real-time CFI, utilizing the programmable frond-end scanner and powerful GPU back-end processor. The parallel architecture and C language development interface of the GPU has enabled easy implementation of parallelism on the CFI algorithm. Experiment on flow phantom had shown that this prototype real-time CFI system could consistently visualize slow flow over a large field on view.