Introduction

It is challenging to capture a good photograph in a low-light environment. Flash can be used to lighten the scene. However, flash could introduce red-eye, uneven illumination, harsh shadows and flash reflection spots etc. Flash also could ruin the natural ambient lighting and make the scene look flat. The most popular compact cameras with small built-in flash placed very close to the optical axis suffer more such kind of artifacts than professional ones. Solutions that do not use flash include prolonging exposure time, enlarging the aperture or using high ISO. These also brighten the image at the expense of image quality, such as introducing blur and noise.

Our Work

Red-eye detection and removal

We have implemented an algorithm to detect and correct red-eye automatically. The main idea of is to combine flash and non-flash images together to the detect red-eye reliably and correct the redness to the true color of human eyes. Given two input images, one captured with flash and the other without flash of the same scene, our algorithm will transfer the eye color from the non-flash image to the flash image if red-eye is detected in the flash image. Following are some results of our method:

f/2.8, 1/60 sec, ISO-200	f/2.8, 1/30 sec, ISO-1600	Corrected Result

The use of two images enable us to recover the true eye color, overcoming the limitations of previous post-processing techniques based on a single defective image. Our method can be implemented on most existing digital cameras that support continuous shooting by modifying only the camera control software. We currently require that the ambient light in the non-flash image not be too low. This minimum light intensity depends on the ISO setting, the aperture size, shutter speed and the camera-specific SNR.

Ambient Image Recovery and Rendering from Flash Photographs

We also proposed a method that combines multiple photographs, taken under varying flash intensities, to recover the intrinsic ambient scene radiance. From this, we can re-render the scene under an arbitrary shutter speed to create a visually pleasant image. We can also simulate the effects of different white-balance settings, as well as different flash intensities.

Ourtechnique can fast recover the ambient scene radiance from a set of flash images.This allows us to capture good pictures in low-light conditions, without requiring slow shutter speed, high ISO, or large aperture. We intend to further pursue our work in several interesting directions: (1) investigate the use of multiple flashes. This will enable us to better deal with shadows and specularities caused by a single flash; (2) incorporate our technique directly into the camera hardware, which will make it extremely convenient to use; and (3)incorporate a suitable red-eye detection and correction technique, so that we can photograph people using flash and avoid red eyes.

This work was funded by the National University of Singapore, project no. R-252-196-112