Re-recording low dynamic range content for high dynamic range display

The invention relates to rerecording low dynamic range content for high dynamic range display. Techniques disclosed herein relate to extending low dynamic range image content (e.g., SDR images) to high dynamic range image content (e.g., HDR images) using a machine learning model (e.g., CNN). The machine learning model may take an image of a low dynamic range as input and may output a plurality of extended images for making the extended image look more natural. The expanded image may be used by an image operator to smooth the color band and to fade the overexposed areas or user interface elements in the expanded image. The extended content (e.g., HDR image content) may then be provided to one or more devices for display or storage. 本公开涉及重新灌录低动态范围内容用于高动态范围显示。本文公开的技术涉及使用机器学习模型(如CNN)将低动态范围的图像内容(如SDR图像)扩展为高动态范围的图像内容(如HDR图像)。机器学习模型可以将低动态范围的图像作为输入，并可以输出多个扩展图，用于使扩展后的图像看起来更自然。扩展图可以由图像算子用来平滑色带，并在扩展后的图像中淡化过度曝光的区域或用户界面元素。然后，扩展的内容(例如，HDR图像内容)可以被提供给一个或更多个设备用于显示或存储。