第2804期:Your phone’s camera isn’t as good as you think(2)

Simply put, to make better digital cameras, you need image sensors with higher numbers of larger photosites. Engineers know this. In fact, it’s basically how they’ve made the best cameras humanity have: giant telescopes that take photos of deep space. But phones don't even have as much sensor space as a standard DSLR camera, let alone the surface area of a massive telescope. In fact, most phone camera sensors are no larger than a pea.简单来说，要制造更好的数码相机，就需要拥有更多、更大的感光元件（photosites）的图像传感器。工程师们对此心知肚明。事实上，人类迄今为止制造出的最强大“相机”——那些拍摄深空的巨型望远镜——正是基于这一原理。然而，手机的传感器面积远小于单反相机，更不用说庞大的望远镜镜面了。实际上，大多数手机相机的传感器都不过豌豆大小。Fortunately, these devices have a technological trick to compensate for their cameras’ tiny size: powerful processors. When you snap a picture on your phone, this pocket-computer starts running complex algorithms, which often begin by secretly taking a string of photos in rapid succession. The algorithms then manipulate these pictures, using math to perfectly align them and identify their best parts before combining the images into one high-quality photo. The end result is an image with less noise, wider dynamic range, and higher resolution than its sensors should be able to achieve.幸运的是，手机拥有一种可以弥补摄像头尺寸不足的技术手段——强大的处理器。当你按下快门拍照时，这个掌上电脑会立即运行复杂的算法，通常会在你毫无察觉的情况下快速连拍多张照片。然后算法会对这些照片进行数学运算，精确对齐每一张图像，挑选出最优部分，再将它们融合成一张高质量的照片。最终生成的图像噪点更少、动态范围更宽、分辨率也远超传感器本身的物理极限。This approach is known as computational photography, and advances here are likely how phone companies will continue to advertise increasingly better cameras without improving their image sensors. Today, these algorithms often leverage machine learning, where phones learn to improve your shots based on patterns found in massive photo databases. For example, night mode prioritizes dynamic range and noise reduction, while portrait mode tells your phone to focus on a central subject and blur the background. Machine learning also allows our phones to do the opposite, unblurring faces to grab quick candid shots. And newer programs can even help you remove unwanted elements altogether.这种方法被称为计算摄影（computational photography）。未来，手机厂商很可能会继续依靠这项技术来宣传“更好的相机”，而不必真正提升传感器硬件。如今，这些算法常常借助机器学习（machine learning），让手机从海量图片数据库中学习如何自动优化你的照片。比如，夜景模式会优先增强动态范围、减少噪点；人像模式则让手机聚焦于主体人物并虚化背景。而机器学习还能反向操作，让手机“去模糊”人脸，从而捕捉自然瞬间。更先进的程序甚至能帮你直接移除照片中不想要的元素。So, with the help of software, even phones with the smallest cameras can snap crisp, detailed photos of loved ones, spectacular views, and of course, lots and lots of food.因此，在强大软件的加持下，即使是最小的手机摄像头，也能拍出清晰细腻的照片——无论是所爱之人的笑容、壮丽的风景，还是那一盘盘令人垂涎的美食。