This repo contains code for UCSD CSE 273 Winter 2022 final project by Zhao Yang [email protected] and Zhengyuan Yang [email protected] It maining consists of a depth-from-strteo algorithm based on energy minimization.

The stereo matching algorithm is described in [1]. Implementation refers to the source code [2] linked with the original paper.

[1] "Kolmogorov and Zabih’s Graph Cuts Stereo Matching Algorithm" by Vladimir Kolmogorov, Pascal Monasse, Pauline Tan (2014) IPOL, http://www.ipol.im/pub/pre/97/

[2] https://github.com/pmonasse/disparity-with-graph-cuts.git

The program assumes rectified image pairs with left image named im0.png and right image named im1.png. The default program to execute .py files is assumed to be Python.

To speed up the computation of the KZ's stereo matching algorithm, one can vertically divide the images into strips and run the algorithm on each pair of strips in parallel. To run KZ's stereo matching algorithm on a pair of images, execute the following command in a Windows command line terminal:
run_disparity.bat imagefolder batchsize overlapsize totalHeight outputfolder numdisparity

• imagefolder - the path to the folder where the rectified image pair is stored
• batchsize - the height of each strip, set to totalHeight / (#CPU cores) to maximize efficiency
• overlapsize - the overlap height of two consecutive strips, should be less than batchsize
• totalHeight - the height of the original image
• outputfolder - the path to the folder where the disparity maps are stored after computation

Use stiching.ipynb to process the resulting disparity strips. Call load_partial() to load disparity strips. Use the same outputfolder, totalHeight, batchsize, and overlapsize parameters as in the last step. Then, call stitch() to stitch disparity stips into a disparity map which aligns with the right image.

Use dof.ipynb to apply depth of field blur effect onto a color image with its disparity map. Call dof_blur() with the following parameters to generate a blurred image.

• color_image - the RGB image to apply DoF blur effect on
• disparity_map - the disparity map of the RGB iamge, should be aligned with the RGB image
• disp_in_focus - the disparity of the focus plane
• levels - the number of disparity levels; increasing levels without changing blurriness also increases the blurriness
• blurriness - how strong the blurriness is if not in focus