CS 143 / Project 1 / Image Filtering and Hybrid Images

Overview

In this project, we create hybrid images by utilizing gaussian filter. The mechanism of hybrid imgaes lies on the frequency domain of perception by human. When hybrid images are close, people tend to observe higher frequency domain. However, when distance is far enough, people now observe lower frequency domain. Therefore, by adding one image only with high frequency and other one with low frequency, we are able to create a hybrid image.

Methods

Matlab's built-in function imfilter was imitated for image filtering. It simply carries out convolution between original image and given filter. Original image was padded with 0 at the edge in order to preserve the size of image after convolution with a filter.

  1. Prepare two images and align them.
  2. Take one image through gaussian filter so that only lower frequency domain remains.
  3. Take another image through gaussian filter and subtract that from the original image, which will leave us higher frequency domain
  4. By adding two filtered images, we create the hybrid image. By varying cutoff frequency, we can get the best result.

Results

Extra Credit