System-Level Optimizations for Mobile User Interfaces: Enhancing Visual Quality, Energy Efficiency, and Responsiveness 

Abstract

Mobile devices have evolved into the primary computing platform for daily interactions, creating an increasing demand for high-quality, responsive, and energy-efficient User Interfaces (UIs). However, delivering a superior user experience (QoE) on mobile platforms entails navigating complex trade-offs between limited hardware resources—such as battery life and thermal constraints—and the growing computational costs of dynamic content rendering. This dissertation addresses these challenges through a series of system-level software optimizations designed to enhance visual fidelity, energy efficiency, and interaction responsiveness across different display technologies and operating system architectures.

First, to address the unique visual challenges of emerging Electrophoretic Displays (EPDs), or e-paper, this research proposes "Ghostbuster," a software-based solution for the ghosting artifacts caused by fast refresh rates. While EPDs are favored for their eye comfort and low power consumption, their slow particle switching speed often leads to visual degradation during dynamic content display. By developing analytical models to predict ghosting occurrences and preemptively adjusting source images to counteract color deviations, this approach significantly improves visual quality on EPDs without requiring hardware modifications.

Second, focusing on the energy efficiency of standard mobile UIs, this dissertation investigates the resource contention between CPU and GPU in heterogeneous mobile systems. As GPUs increasingly handle general-purpose workloads, UI rendering tasks often face blocking, leading to frame drops. We propose a QoS-aware task migration strategy that dynamically offloads specific rendering tasks from the GPU to the CPU. Coupled with an intelligent frequency scaling mechanism that balances power and performance, this method reduces total system power consumption by 29% while maintaining a smooth frame rate.

Third, to tackle the issue of UI responsiveness in modern mobile operating systems like Android, we address the bottleneck caused by the unified handling of input events and rendering on the main thread. Heavy rendering workloads frequently block user input processing, causing perceptible lags. This work restructures the UI rendering workflow by decoupling event handling from rendering tasks. By implementing tailored techniques based on whether event results require immediate screen updates, we achieve significant reductions in both average and worst-case response times, thereby ensuring a fluid interactive experience.

Collectively, these contributions demonstrate that strategic, system-level software interventions can effectively overcome hardware limitations. This dissertation provides a comprehensive framework for optimizing the mobile UI rendering pipeline, offering scalable solutions that enhance visual performance, prolong battery life, and minimize interaction latency.

Date of Award	15 Apr 2026
Original language	English
Awarding Institution	City University of Hong Kong
Supervisor	Nan GUAN (Supervisor)