Spandan Dutta(spandan)

1. Algorithm

first classify the image into grayscale or rgb image

if the image is grayscale we proceed as follows:

(1) We use a gaussian filter and call my_imfilter using this filter on image1. This gives us back an image1 with only the low frequencies.

(2) We use a gaussian filter on image2 as well when calling my_imfilter. This will give us the low frequency component of image2. Subtract this from the original to get the high frequency component of image2.

(3) We finally add the results of step1 and step2 to get the hybrid image.

if the image is rgb then we proceed as follows:

(11) First we extract the three components of an rgb image.
(22) We treat each component separately. We do steps (1),(2) and (3) of the gray scale case, for each component.
(33) Finally ,create a new rgb image with these three new components as calculated in step (22).

my_imfilter

Here we are trying to implement the convolution operation between the filter and the image. We first create a ouput image having the same resolution as either one of the images, with each pixel having zero intensity. We also pad the original image with rows and columns containing 0s based on the size of the filter. [padder.m has the implementation.] To do the task in the previous mentioned sentence, we compute the center of the filter matrix. Finally we compute the convolution at the every pixel belonging to the original image and store that value in the corresponding output matrix. [my_conv.m contains the necessary implementation of this.]

2. Results in a table