最邻近插值
和matlab自带函数完全一样。
function nI = Nearest_Neighbor(I, ratio)
% 将原图像 I 以 1:ratio 的倍率缩放得到目标图像 nI
% 插值方法为 最临近插值
[n m mm] = size(I);
new_n = round(n * ratio);
new_m = round(m * ratio);
nI = zeros(new_n, new_m, mm, 'uint8');
for i = 1 : new_n
for j = 1 : new_m
x = (j-0.5) / new_m * m + 0.5;
y = (i-0.5) / new_n * n + 0.5;
nI(i, j, :) = I(round(y), round(x), :);
end
end
双线性插值
和matlab自带函数有最多为1的误差,感觉是由于matlab进行了优化,没有进行浮点数运算导致的误差。
function nI = Bilinear_Interpolation(I, ratio)
% 将原图像 I 以 1:ratio 的倍率缩放得到目标图像 nI
% 插值方法为 双线性插值
[n m mm] = size(I);
new_n = round(n * ratio);
new_m = round(m * ratio);
nI = zeros(new_n, new_m, mm, 'uint8');
I = [I(1,:,:); I; I(end,:,:)];
I = [I(:,1,:), I, I(:,end,:)];
I = double(I);
for i = 1 : new_n
for j = 1 : new_m
x = (j-0.5) / new_m * m + 1.5;
y = (i-0.5) / new_n * n + 1.5;
r1 = floor(y); r2 = ceil(y);
c1 = floor(x); c2 = ceil(x);
Q11 = I(r1, c1, :); Q12 = I(r1, c2, :);
Q21 = I(r2, c1, :); Q22 = I(r2, c2, :);
R1 = Q11 + (Q12-Q11)*(x-c1);
R2 = Q21 + (Q22-Q21)*(x-c1);
nI(i, j, :) = uint8(R1 + (R2-R1)*(y-r1));
end
end