The Retinex Decomposition Model for X-ray Coronary Angiographic Sequences

Since the coronary angiographic vessel images take the low image contrast, the severe artefact, and blurred vascular structures, the complex background structure makes the image quality extremely poor. These shortcomings make it much more difficult to analyze angiographic vessels than the general vessel, such as OCT fundus vessels. To obtain high quality coronary angiographic images for post-processing, we propose the Retinex decomposition model. In this model, the foreground vessel and background part of the angiographic data can be regarded as the image reflectance and illumination, respectively. By using y direction gradient for the prior representation of coronary angiographic sequences, the feature matrix of coronary angiographic sequences in the log-transform domain is decomposed. Different from existing works, the proposed model can remove artefacts with the good visual performance. Moreover, the proposed model can perverse the vessel structure to some extent. Synthetic experiment experiment demonstrate the effectiveness of the proposed model compared to state-of-the-art methods both subjective and objective assessments.