What are the best methods for image fusion in medical imaging applications?

1 Image fusion methods

Image fusion can be done in different ways, depending on the type and level of information to be merged. Pixel-level fusion is a common method, which involves operating on the raw data of the images, such as intensity, color, or texture. This approach is fast and simple, but may not preserve the spatial and spectral features of the images. Feature-level fusion extracts meaningful features from the source images, such as edges, contours, or regions of interest. This method is more complex and computationally intensive, but can enhance the contrast and detail of the images. The most sophisticated method is decision-level fusion, which uses high-level information from the source images, such as labels, classifications, or decisions. This approach is application-specific, but can provide the best interpretation and understanding of the images.

2 Image fusion quality assessment

Image fusion quality assessment is the process of evaluating the performance and effectiveness of the image fusion methods. This process aims to answer questions such as how well the fused image preserves relevant information from the source images, reduces noise and artifacts, and improves visual perception and interpretation. There are two main approaches to image fusion quality assessment: subjective and objective.

3 Subjective quality assessment

Subjective quality assessment relies on human judgment and feedback in order to rate or compare fused images according to criteria such as clarity, contrast, or detail. This method is useful for capturing the human perception and preference of the images, but it can be time-consuming and expensive to conduct. Additionally, it is subject to human factors such as fatigue, bias, or preference which can make it difficult to quantify and compare the results.

4 Objective quality assessment

Objective quality assessment is a mathematical process that computes numerical values to evaluate the quality of fused images according to criteria such as information content, distortion, or similarity. This method is useful for providing a fast and consistent evaluation, however, it can be difficult to define and measure quality criteria for various applications and scenarios. Additionally, it is difficult to find a universal metric that captures all aspects of quality for different fusion methods and levels, as well as to correlate the objective metrics with subjective ratings.

5 Hybrid quality assessment

Hybrid quality assessment is a combination of subjective and objective approaches, which involves using both human feedback and mathematical models to assess the quality of the fused images. This method is beneficial for overcoming some of the drawbacks of each approach, however, it requires more resources and coordination. Examples of hybrid quality assessment include user-guided evaluation, which allows the user to adjust parameters or weights to optimize the quality; learning-based evaluation, which uses machine learning techniques to learn the relationship between objective metrics and subjective ratings; and multi-criteria evaluation, which uses multiple objective metrics and subjective ratings to evaluate the quality from different perspectives.