ABOUT HDR IMAGES.

HDR (High Dynamic Range) images are images with a significant difference between light and dark areas. This means that the range of colors and grey levels that the pixels of these images display is wider than that which can be captured by a conventional camera with one exposure. This image can also not be displayed on a conventional computer monitor, being a cathode ray tube (CRT) or a liquid crystal display (LCD).

In the real world, the luminosity can vary from 10 exp-3 cd/m² (light generated by a starlight at night) 10 exp+5 cd/m² (sunlight in full day), and the contrast of a scene can be in a ratio of 1 / 100 000 (ratio of light intensity between black and white). A standard image on a computer screen typically offers a contrast ranging from 1/50 to 1/500. These types of images are called LDR (Low Dynamic Range) images.

The problems linked to LDR images are :

- They do not preserve details (compression).
- You cannot see both very luminous objets and very dark objects at the same time, and also see the details in both cases.
- You cannot lighten or darken an image.

Digital cameras now allow you to shoot pictures with several exposures simultaneously (« bracketing ») by varying the shutter speed. Typically, to create a HDR image, one will take the same picture with from 3 to n (for example, n=9) different exposures. HDR images can therefore be created quite easily. They will be encoded on 16 or 32 bits per channel using a floating-point representation. The information, which is encoded this way, corresponds to actual luminance values. As an analogy, LDR images are encoded on 8 or 16 bits in integer values. The information, which is encoded this way typically for each one of the red, green and blue channels, correspond to the values to be displayed on the screen.