I\'m trying to create a function the scales an image based on a value (scale_zoom) for a homework assignment. I don\'t want to use the MATLAB built in function resize()
I had previously answered a question about scaling images using nearest-neighbor interpolation, and much of the code and explanations I used there apply here. The main difference is the final interpolation step. Here's how you can write your function, using INTERP2 and generalizing for a 2-D grayscale or 3-D RGB image of any type:
function pic_new = scale_image(pic,scale_zoom)
oldSize = size(pic); %# Old image size
newSize = max(floor(scale_zoom.*oldSize(1:2)),1); %# New image size
newX = ((1:newSize(2))-0.5)./scale_zoom+0.5; %# New image pixel X coordinates
newY = ((1:newSize(1))-0.5)./scale_zoom+0.5; %# New image pixel Y coordinates
oldClass = class(pic); %# Original image type
pic = double(pic); %# Convert image to double precision for interpolation
if numel(oldSize) == 2 %# Interpolate grayscale image
pic_new = interp2(pic,newX,newY(:),'cubic');
else %# Interpolate RGB image
pic_new = zeros([newSize 3]); %# Initialize new image
pic_new(:,:,1) = interp2(pic(:,:,1),newX,newY(:),'cubic'); %# Red plane
pic_new(:,:,2) = interp2(pic(:,:,2),newX,newY(:),'cubic'); %# Green plane
pic_new(:,:,3) = interp2(pic(:,:,3),newX,newY(:),'cubic'); %# Blue plane
end
pic_new = cast(pic_new,oldClass); %# Convert back to original image type
end
And you can test it as follows:
img = imread('peppers.png'); %# Load the sample peppers image
newImage = scale_image(img,0.3); %# Scale it to 30%