Effective data visualizations are essential for communicating complex information clearly and intuitively. However, many charts suffer from visual shortcomings such as poor text legibility, color schemes that are inaccessible to users with color vision deficiencies (CVD), and suboptimal visual attention allocation, all of which hinder interpretability. This thesis extends the Aalto Interface Metrics (AIM) repository by implementing and integrating five novel chart-specific evaluation metrics to address these challenges. The metrics include: (1) Text Legibility using OCR-based scoring, (2) Color Blindness Accessibility through perceptual simulations of protanopia, deuteranopia, and tritanopia, and (3) three state-of-the-art saliency models—VisSalFormer, SUM, and ScannerDeeply—designed to predict and evaluate visual attention in static charts. The methodology involves modular integration of these Python-based metrics into the existing AIM pipeline, ensuring compatibility with standardized input/output formats and maintaining scalability. The SUM model was retrained using a chart-focused saliency dataset provided by the original authors, enabling better alignment with visualizationspecific contexts. Each metric was validated using both synthetic and real-world charts, with results visualized through saliency heatmaps, simulated CVD renderings, and quantitative accessibility or saliency scores. Beyond technical implementation, this thesis also contributes a structured literature review of evaluation metrics in information visualization, positioning the developed framework within a broader research landscape. The modular design enables future extensions, such as incorporating domain-specific metrics or integrating real-time user feedback mechanisms. Together, these contributions lay the groundwork for a more comprehensive, data-driven approach to visualization quality assessment—bridging perceptual theory and computational analysis to support more inclusive and interpretable visual design.