Combining spectral total variation with dynamic threshold neural P systems for medical image fusion

Synthesis of medical images is one of the indispensable tasks today because of its applications in clinical diagnosis. Composite images often suffer from problems such as poor contrast, loss of detail, and low light intensity. The reason for the above problem is that the input image is of poor quality, and the fusion rules are not really effective. In this paper, we propose a new image synthesis model to simultaneously solve the problems mentioned above. Firstly, the input image is enhanced because the input image's quality significantly affects the fusion image's quality. Next, the Spectral total variation (STV) method is utilized to decompose input images into a base layer and a series of detail layers. An adaptive rule based on the Chameleon Swarm Algorithm (CSA) algorithm is proposed for the synthesis of the base layers. This rule ensures that the synthesized image has good quality in terms of brightness and contrast. To ensure that the details are preserved in the synthesized image, we propose an effective fusion rule for detail layers based on the Dynamic threshold neural P systems (DTNPS). Finally, the base and detail layers that have been composited are summed together to create the composite image. Six evaluation indexes, seven state-of-the-art image synthesis algorithms, and 132 medical images were used to evaluate. The results show that our image synthesis model is more efficient than the current latest image synthesis methods.