System and method for generating bokeh image for dslr quality depth-of-field rendering and refinement and training method for the same

An electronic device comprising:

• at least one memory configured to store multiple reference images of a scene;
• and
• at least one processing device configured to:
• process the reference images using a first convolutional neural network (CNN) to generate a confidence map and a disparity map;
• generate an initial Bokeh image based on the disparity map and the reference images using a depth-of-field (DoF) renderer;
• and
• refine the initial Bokeh image using a second CNN to generate a refined Bokeh image, wherein the second CNN is configured to use the confidence map, the disparity map, and the reference images to generate the refined Bokeh image.

A method comprising:

• obtaining (902), using at least one processor of an electronic device, training data comprising multiple reference images (611, 612), a ground truth disparity map (613), and a ground truth Bokeh image (614);
• training (904), using the at least one processor, a first convolutional neural network (CNN) (210) using the reference images and the ground truth disparity map;
• and
• training (906), using the at least one processor, a second CNN (230) using the reference images, an initial Bokeh image (222), a confidence map (214) generated during the training of the first CNN, and a predicted disparity map (212) generated during the training of the first CNN.