Mobile technology has progressed beyond the scope of communication, increasingly influencing sectors such as education, health, and finance. For the 15% of the global population living with disabilities, accessibility is arguably the most crucial software quality attribute. Leading mobile operating systems, including iOS and Android, offer various built-in assistive services to enhance accessibility for users with disabilities. Such services include screen readers (for people with visual impairments) and switches (for those with mobility impairments), providing disabled individuals with tools to use phones effectively and accomplish tasks that might be otherwise challenging or not possible. App accessibility relies on following guidelines, best practices, and extensive testing to confirm compatibility with assistive services. Failure to comply with these requirements can lead to accessibility issues and barriers for users.

This dissertation aims to enhance mobile app accessibility by initially conducting a large-scale empirical study involving apps, developers, and users to determine the prevalence, categories, and characteristics of accessibility issues, along with development practices that might have contributed to these issues. Next, driven by insights from the study, the research focuses on improving app accessibility for low-vision users, especially those relying on the Text Scaling Assistive Service (TSAS). This is achieved by proposing practical methods and techniques to detect, localize, and automatically repair text accessibility issues stemming from incompatibility between apps and TSAS. The dissertation introduces AccessiText, an automated tool designed to accurately detect text accessibility issues by analyzing UI screenshots and metadata information collected through dynamic analysis. AccessiText employs various heuristics based on distinct types of text accessibility issues, discovered by examining user-reported feedback in Play Store reviews and Twitter data. Furthermore, this dissertation presents Artex, a search-based automatic repair technique utilizing a genetic algorithm to localize and automatically repair text accessibility issues, while minimizing layout distortion and preserving the app layout's consistency.

In our evaluation of the proposed techniques and tools, we conducted experiments and user studies on real-world commercial applications. The findings demonstrated the effectiveness, efficiency, and usefulness of these techniques in resolving text accessibility issues.