CHIMERA: Adaptive Cache Injection and Semantic Anchor Prompting for Zero-shot Image Morphing with Morphing-oriented Metrics

A new study introduces Prompt-Conditioned Intervention (PCI) to analyze the dynamics of concept formation in diffusion models, focusing on how noise transitions into specific concepts during the image generation process. This framework evaluates Concept Insertion Success (CIS), which measures the likelihood that an inserted concept is preserved in the final output, providing insights into the controllability and reliability of these models.

A novel framework called self-refining diffusion has been introduced to enhance image generation quality in diffusion models by detecting artifacts and unrealistic regions. This framework utilizes explainable AI-based flaw activation maps (FAMs) to identify and improve flawed areas during the image synthesis process, achieving significant performance improvements across various datasets.

A new method for Positron Emission Tomography (PET) image reconstruction has been proposed, utilizing diffusion models to enhance image quality while addressing challenges such as tracer-specific contrast variability and high computational demands. This approach employs an anatomical prior-guided framework that alternates between diffusion sampling and model fine-tuning, resulting in improved image reconstruction from various PET tracers.