Images color rendering accuracy analysis after applying noise reduction models

Image restoration approaches are widely-used. Frequently, the tasks of image resolution enhancement and image noise canceling on are solved using neural networks. After applying an algorithm or neural network model to an image, the result may contain distorted colors during the restoration process due to information losses. Distortion level is hard to estimate because clear images marked as ground-truth samples may still contain noise components. For image enhancement tasks, it is extremely important to save the original colors when they are transformed. To solve this problem it is necessary to use special devices which allow us to calculate color rendering quality of the obtained image. In this work we estimated the level of color rendering preservation for the results of modern neural network models for image noise reduction.