CS 194-26 Project #2
Fun with Filters and Frequencies!
This project involved creating various filters and applying them to images to create different results. The process involved creating the filter and then applying the images. There are some intermediate steps including separting the color channels (in some cases) or also aligning the image using provided starter code.
1.1 Finite Difference Operator
This part included taking the finite difference operator and applying it to the cameraman image to showcase the partial derivative in x and y. Gradient was calculated by combining the partial derivtives of an image into a single vector.
x derivative
y derivative
gradient
edge (0.25 threshold)
1.2 Derivative of Gaussian (DoG) Filter
This part included creating a Gaussian filter and convolving the filter with the derivative in x and y to showcase the resulting DoG filters as images.
G convolved with dx
G convolved with dy
image convlved with G convolved with dx
image convolved with G convolved with dy
gradient
egde (threshold 0.059)
I noitced that my gradient looks vastly different from other submissions. I wasn’t too sure what the problem was and couldn’t figure it out. However, I did notice that compared to part 1.1, I had to lower the threshold by a large amount for the binarized gradient image.
2.1 Image “Sharpening”
In this part, we sharpened an image using the unsharp mask filter. We obtain the high frequencies of the image by subtracting the blurred version from the original image.
original taj image
blurred taj image
sharpened taj image with alpha = 1
sharpened taj image with alpha = 2
original image of almansor park
blurred park image
sharpened park image with alpha = 1
2.2 Hybrid Images
This part involved creating hybrid images out of two images. The idea is that the image will look differnt depending on how close you are while viewing it. The implementation involved creating a low-pass and high-pass filter, filtering the respective images with their filters, and then combining the two images!
picture of derek
picture of cat nutmeg
hybrid picture with nutmeg being in high sf
my dog, kacey!! <3 :D <3
aang from avatar the last airbender
hybrid image (FAILURE)
sam smith
stan smith
hybrid image
2.3 Gaussian and Laplacian Stacks
This part involved creating the Gaussian and Laplacian stacks that would be later be used for multiresolution blending. To create the Gaussian stack, we just applied the Gaussian filter multiple times down the stack. For the Laplacian stack, we took the values from our Gaussian stack.
First Level
2.4 Multiresolution Blending (a.k.a. the oraple!)
apple
orange
oraple
beach day
beach night
beach day/night
CS 194-26 Project #2
Fun with Filters and Frequencies!
This project involved creating various filters and applying them to images to create different results. The process involved creating the filter and then applying the images. There are some intermediate steps including separting the color channels (in some cases) or also aligning the image using provided starter code.
1.1 Finite Difference Operator
This part included taking the finite difference operator and applying it to the cameraman image to showcase the partial derivative in x and y. Gradient was calculated by combining the partial derivtives of an image into a single vector.
x derivative
y derivative
gradient
edge (0.25 threshold)
1.2 Derivative of Gaussian (DoG) Filter
This part included creating a Gaussian filter and convolving the filter with the derivative in x and y to showcase the resulting DoG filters as images.
G convolved with dx
G convolved with dy
image convlved with G convolved with dx
image convolved with G convolved with dy
gradient
egde (threshold 0.059)
I noitced that my gradient looks vastly different from other submissions. I wasn’t too sure what the problem was and couldn’t figure it out. However, I did notice that compared to part 1.1, I had to lower the threshold by a large amount for the binarized gradient image.
2.1 Image “Sharpening”
In this part, we sharpened an image using the unsharp mask filter. We obtain the high frequencies of the image by subtracting the blurred version from the original image.
original taj image
blurred taj image
sharpened taj image with alpha = 1
sharpened taj image with alpha = 2
original image of almansor park
blurred park image
sharpened park image with alpha = 1
2.2 Hybrid Images
This part involved creating hybrid images out of two images. The idea is that the image will look differnt depending on how close you are while viewing it. The implementation involved creating a low-pass and high-pass filter, filtering the respective images with their filters, and then combining the two images!
picture of derek
picture of cat nutmeg
hybrid picture with nutmeg being in high sf
my dog, kacey!! <3 :D <3
aang from avatar the last airbender
hybrid image (FAILURE)
sam smith
stan smith
hybrid image
2.3 Gaussian and Laplacian Stacks
This part involved creating the Gaussian and Laplacian stacks that would be later be used for multiresolution blending. To create the Gaussian stack, we just applied the Gaussian filter multiple times down the stack. For the Laplacian stack, we took the values from our Gaussian stack.
First Level
2.4 Multiresolution Blending (a.k.a. the oraple!)
apple
orange
oraple
beach day
beach night
beach day/night