gancheekim / ntu-vfx2023-hw1

NTU VFX2023 Homework 1: High Dynamic Range Imaging

Description:

website: https://www.csie.ntu.edu.tw/~cyy/courses/vfx/21spring/assignments/proj1/

TL;DR:

1. Taking photographs.
   • Take a series of photographs for a scene under different exposures. (by changing shutter speed)
2. Write a program to assemble an HDR image.
   • we implemented the Paul Debevec's method. (Recovering High Dynamic Range Radiance Maps from Photographs, SIGGRAPH 1997.)
3. Develop your radiance map using tone mapping.
   • we implemented Reinhard's method. (Photographics Tone Reproduction for Digital Images, SIGGRAPH 2002.)
4. Bonuses:
   • we implemented the MTB image alignment method for HDR imaging.

Member:

• 電信所碩一 顏子鈞
• 電信所碩一 陳詠源

Dependencies:

please install them using pip install -r requirements.txt. packages used in our code:

• numpy, pillow, opencv-python, matplotlib

Execution:

Generate HDR Image and Tonemapped LDR Image:

$ cd code
$ python main.py --dataset <name_of_dataset> --dataset_info <path_to_textfile> --N <num_of_samplepoints_per_image> --tm_key <key_to_control_tonemapping>

Note:

• to disable MTB image alignment, use --disable_mtb flag
• choices of <name_of_dataset>: memorial, ntu_sample1, ntu_sample2, ntu_sample3, ntu_sample4, default = ntu_sample1
• <num_of_samplepoints_per_image> is the sampling point used for calculating Debevec HDR algorithm, default value = 40
• <key_to_control_tonemapping> is the key parameter of Reinhard's photographic tonemapping algorithm, default value = 0.5
• output files are all save at "data/output/" directory, including HDR irradiance map, camera response curve, tonemapped LDR image (global and local operation).

Results:

ntu_sample1:

• original captured images:

• calculated irradiance map:

• reconstructed HDR images:

ntu_sample3:

• original captured images:

• calculated irradiance map:

• reconstructed HDR images: