High Dynamic Range Imaging

CS 194-26 Final Project 3 - Fall 2020

Glenn Wysen and Morgan Nanez

doorway
Fig. 1 - Over exposed source image.
doorway
Fig. 2 - Radiance map taken over all exposures of initial image.

Radiance Maps

The first step to recovering an HDR image was to find radiance maps for images which represent the overall radiance from a series of images taken of the same target with different exposures. This process involved finding a relation between pixel values (range 0-255) to exposure levels (relationship shown in Fig. 3). Higher values are shown in the green/yellow while lower values are blue. Another visualization of this mapping is shown in Fig. 4 below which shows an example curve for one of the images mapping pixel values to log(intensities).

doorway
Fig. 4 - A plot showing the relationship between pixel value and intensity for a random sample image.
doorway
Fig. 3 - Radiance map taken over all exposures of initial image for the red channel.

Bilateral Filtering

Another step in the process was performing bilateral filtering and recovering the high detail parts of the radiance maps. A visualization of these two filters for the red channel of the chapel is shown in Fig. 5 and Fig. 6 below.

doorway
Fig. 5 - Bilateral filter applied to red channel radiance image.
doorway
Fig. 6 - Detail map for the same image highlighting the parts where there is high radiance.

Example Results

Below are some additional results of running the whole algorithm over some sample inputs.

doorway
Fig. 7 - Long exposure.
doorway
Fig. 9 - Short exposure.
doorway
Fig. 8 - Medium exposure.
doorway
Fig. 10 - Very short exposure.
doorway
Fig. 11 - HDR image.
doorway
Fig. 12 - Long exposure.
doorway
Fig. 14 - Short exposure.
doorway
Fig. 13 - Medium exposure.
doorway
Fig. 15 - Very short exposure.
doorway
Fig. 16 - HDR image.
doorway
Fig. 17 - Long exposure.
doorway
Fig. 19 - Short exposure.
doorway
Fig. 18 - Medium exposure.
doorway
Fig. 20 - Very short exposure.
doorway
Fig. 21 - HDR image.