For the Bonus part (i.e. finding horizontal seams to reduce the image both in height and width), we implemented one primary function called flip_RGB_image() which, as the name suggests, rotates the image by 90 degrees. Thanks to this function we were able to apply the same processes we as before but this time with a rotated image. This allowed us to find the usual vertical seams for the flipped images which corresponded to the horizontal seams on the original image.
To implement this process we split it in two parts: the first one in which we identified the vertical seams on the original image and the second where we did the same thing but with the flipped image. By rotating again the image and combining it with the first one, we obtained a final image with both horizontal and vertical seams highlighted.
The process was very similar for removing these seams to scale the image. Again, we split the process in two parts: during the first one we found and removed the vertical seams on the original image and then, in the second part, we flipped the resulting image and applied the same process again to find and remove horizontal seams. Finally we flipped again the image to restore its former orientation.
In the extension.cpp file we added also some function which were meant for testing, for example a function that outputs the resulting image after applying only the sobelX or sobelY filter to it. Furthermore we added some functions to test different kernels such as the Gaussian blurring 5x5 kernel or one that is supposed to sharpen the edges of the image, adding contrast. Finally we also wrote a function to check if two images are equal pixel by pixel in order to confront our outputs to the expected ones that were given in the .zip file.
In the files extension.cpp and extension.h you will find quite some code that has been commented out. This is because, not satisfied with the efficiency of our program, we tried to work out a better (faster) solution but didn't have enough time to obtain any valuable result.