Computer Graphics 2024 - Project
Published:
Blog posts
2025
2023
Bayesian Occam’s Razor
Published:
In the 14th century, an English philosopher and theologian, named Willam of Ockham, stated that “Entia non sunt multiplicanda praeter necessitatem”, which translates as Entities must not be multiplied beyond necessity . This principle is also known as Occam’s razor. If there are multiple hypotheses compatible with an observed dataset, Occam’s razor then advises us to choose the simplest. Although the conclusion seems hasty, it can be explained and understood as a consequence of Bayesian inference, a method of updating beliefs based on evidence.
2022
Variational Methods and 3D Reconstruction
Published:
There are numerous methods for reconstructing a real object into the mesh, such as voxel carving, which independently processes the input images or structure from motion. However, in this blog, we would like to introduce to you another solution for this ill-posed inverse computer vision problem. This method is a volumetric approach, where each voxel is assigned two probability values for being in or out of the 3D object. Let’s start.
Variational Methods and 3D Tracking
Published:
This blog presents you with a lightweight real-time segmentation and pose tracking method which only uses only a monocular RGB camera and can run with multiple objects, and is robust to occlusion. You can watch the demonstration first to have a sense. The whole process is just mathematics which makes the output of each step predictable and allows us to have insight for further improvement. Real-Time Monocular Segmentation and Pose Tracking.
Variational Methods and Image Segmentation (Part 3)
Published:
This blog will show you the general version of the active contours model via statistical perspective and how to enhance the performance of the method.
Variational Methods and Image Segmentation (Part 2)
Published:
After finishing Snakes problem in part 1, today, we will get into an improvement of it which is called Active Contours Without Edges. The reason it has “without edges” is that the model doesn’t use the image gradient information of the input image. You can also read the original version at here.
Variational Methods and Image Segmentation (Part 1)
Published:
Convolution neural networks usually appear in segmentation problems because of their high adaptation to many datasets and high performance. However, in return, they require ground truth data to learn and perform specific tasks, and without ground truth, their results are really poor. Today, I will introduce to you “old” segmentation methods, but they can be applied to several certain problems in the absence of datasets. This is called Snakes: Active Contours Models.
Variational Methods and Image Denoising
Published:
Variational methods are really powerful and have a variety of applications, such as 2D segmentation and 3D reconstruction. However, today I will present to you one of their interesting applications: Image Denoising.
2020
Video Stabilization
Published:
Video stabilization is a process that aims to reduce the vibration and jitter inside videos.
Harris Corner Detection
Published:
A Corner is a point whose local neighborhood stands in two dominant and different edge directions. In other words, a corner can be interpreted as the junction of two edges, where an edge is a sudden change in image brightness. Corners are the important features in the image, and they are generally termed as interest points that are invariant to translation, rotation, and illumination.
Optical Flow
Published:
Optical flow is the pattern of appearance of objects’ motion in an image, between two successive frames caused by the movement of objects or the camera. It is a 2D vector field, where each vector represents a displacement or movement of features from a first frame to a second frame (see the below image).