Project 1: Image Filtering and Hybrid Images

Hybrid Image

Filtering Algorithm

First of all I pad the input image with zeros on both the dimensions. The number of padded zeros on each dimension is half the size of the filter in that dimension.


			x = floor(size(filter,1)/2);
			y = floor(size(filter,2)/2);
			image = padarray(image,[x,y]);

Next I iterate over all the 3 dimensions of the input image and element-wise multiply a block of the image of size [2x+1,2y+1] with the filter.


			for i = 1:Size(1)
				for j = 1:Size(2)
					for k = 1:Size(3)
						pos = image(i:i+2*x,j:j+2*y,k) .* filter;
						output(i,j,k) = sum(pos(:));
					end
				end
			end

Image Filtering Results

Blur Image Large Blur Image Sobel Image Laplacian Image High Pass Image
Blur Image Large Blur Image Sobel Image Laplacian Image High Pass Image

Hybrid Image Results

Low Frequency High Frequency Hybrid Image
Low Frequency High Frequency Hybrid Image

Analysis

I'm using 2D filtering as I found it to be a bit faster than 2 1D filtering, mostly because we have small and less number of images. I found cut-off frequency = 5 to give better results for the image pairs I chose. I'm filtering the 3 color channels using a for loop and storing the values in a 3D array instead of doing the filtering separately and concatenating the 3 channels.