Image Recognition has been a widely studied topic recently. Researchers and companies are trying to find robust solutions to recognize images (advertisement flyers, movie posters, cards …) and objects using remote servers or directly on the devices. At iplimage, we tried and implemented a few image recognition engines lately, and I would like to share some results.

# [geom] Computing an homography between two views

This article describes how to construct a geometric transformation between two images representing two views of an object.

# [geom] Choosing an internal coordinate system

There are many ways to express affine transformations. In this post are discussed several possibilities and the choice I’m used to make when implementing an application that requires 3D transformations.

# XZIMG Face Detection and Face Tracking Technology

**XZIMG Face Tracking engine** is a product developed by XZIMG Research that addresses the **Augmented Reality** market. Using its components you will be able to create experiences based on the Face Tracking technology.

# “Hybrid” Image (Framed Image) Detection

I’m trying to implement an “hybrid” image tracker. This technique has the objective to detect certain type of images to enable augmented reality. Here are early results!

# [video] – Kalman Filter

Here is given an example on how to use the well known **Kalman filter** to filter orientation values and thus reduce jitters when tracking objects recursively.

# [math] – Kalman Filter

For my own comprehension (i hope that it can help others too), here is a short post about the very famous Kalman filter. In 1960, R. E. Kalman published his paper presenting a recursive solution for the discrete-data filtering problem. Continue reading

# [math] – LU Decomposition

LU Decomposition consists in rewriting a square matrix $A$ into an unit lower triangular matrix $L$ and an upper triangular matrix $U$. Continue reading

# Create a unity project that opens a webcam

Unity3d is now widely used for creating multi-platform games and applications that uses 3d. This post summarily describes how to create a new unity project using a direct show based video capture module i created.

# [math] – The P3P (Perspective-Three-Point) Principle

As said and shown before, the P3P equations are not easy to solve. As an example, the solution provided in the opencv library does not compute fully satisfactory results. This article has the objective to summarize the best strategy to implement the P3P approach.