The goal of this assignment is to have a better understanding of the importance of high and low frequences in an image. This will be accomplished by two practical examples: the hybrid images experiment and the blending with Laplacian Sequences. The results (obtained over my own pictures, unless otherwise noted) and a critical evaluation are shown bellow.
Part 0: Warm Up
Here the goal was to sharp a given image by adding the image details (image - gaussianBlur(image)) to itself.
Part 1: Hybrid Images
Hybrid image for the provided image pair.
Part 1: Best Result & Supplementar Credits
The color seems to improve the results when used in the low-frequency image. Otherwise, when the scene is seen from far away (where only the low-pass frequency should be visible) the high-frequence image will still be easy to see since the colors will be enhancing the edges locations of the high-frequence image.
Part 2: Gaussian and Laplacian sequences
Sequences for the "Lincoln et Gala" image.
Gaussian Sequence.
Laplacian Sequence.
Sequences for the "Ferreetle" image.
Gaussian Sequence.
Laplacian Sequence.
Part 3: Multiresolution Mixture
In this part of the exercise we were asked to mix two images by calculating the Laplacian sequences (L_ImA(i) and L_ImB(i)) of the images and the Gaussian sequences (G_Mask(i)) of the mask. To produce the blended image, we can sum the images by the following equation: (G_Mask(i)*L_ImA(i)) + (1-G_Mask(i)*L_ImB(i)).