c00renut / cv-textureanalysis

Code to perform Texture Synthesis by Non-Parametric Sampling as proposed by Efros and Leung, ICCV99

Computer Vision - Texture Synthesis

This is the first part of the project where the goal is to implement and evaluate the idea of Texture Synthesis by Non-Parametric Sampling as proposed by Efros and Leung, ICCV99.

The paper can be found at this link : http://graphics.cs.cmu.edu/people/efros/research/NPS/efros-iccv99.pdf

Getting Started

We are given a pseudo code of the implementation of Non-Parametric Sampling algorithm which we need to run on the sample gif images to synthesize a 200x200 pixel image.

Example image :


We repeat this for various WindowSize parameters and evaluate the performance and quality of the synthesized image.

The pseudocode for the algorithm : http://graphics.cs.cmu.edu/people/efros/research/NPS/alg.html

We plot synthesize images for various window sizes hence resulting in synthesised images becoming better qualitatively however taking a longer time to be synthesized. The results for synthesized images for different seed images are as follows:

Texture 1 :

Texture 2 :

Texture 3 :

Texture 4 :

Texture 5 :