How do you find and remove a seam when doing seam carving in Python for image processing?

I am implementing the seam carving algorithm in Python, which is used to resize and manipulate images.

See the following links for an explanation of the seam carving algorithm:
http://cs.brown.edu/courses/cs129/results/proj3/taox/
http://www.cs.princeton.edu/courses/archive/spring13/cos226/assignments/seamCarving.html

The dual_gradient_energy function computes the energy of each pixel of the image.

I am trying to find the horizontal seam of the image.

The function is supposed to return an array of H (height) integers. For each row, it should return the column of the seam.

I am unable to get the find_seam function to work. Instead of returning pixels in integers, the values being returned in the array are in scientific notation. I made sure to check boundary cases so that it does not fill an array with an invalid/nonexistent element.

Plot_seam should plot the seam on the image and highlight it in red. I have it implemented already, but I have not been able to test it since there is no seam to plot. Once the find_seam function is fixed, then a seam can be plotted.

Remove_seam function modifies the image in-place and returns a W-1 (width) x H x 3 (height) slice. In essence, it removes the seam from the image. I am having trouble implementing this because the array indices seem to be out of bounds.

I have provided an image (someimage.png) to use for testing the functions.

I am using Python 2.7.

Required Python libraries (for Windows 32-bit):

numpy-MKL-1.9.0.win32-py2.7
scipy-0.14.0-win32-superpack-python2.7
scikit-image-0.10.1.win32-py2.7
matplotlib-1.4.0.win32-python2.7
python-dateutil-2.2.win32-py2.7
pyparsing-2.0.3.win32-py2.7

What is wrong with my find_seam and remove_seam functions? How can I fix my code?

Note: You may need to close the graph that is generated when you run the source code in order to see the find_seam, plot_seam, and remove_seam function calls.

See my code below:

import numpy
import scipy.misc as scm
from pylab import *
from scipy import ndimage
from skimage import img_as_float, filter


def dual_gradient_energy(img):
    R = img[:, :, 0]
    G = img[:, :, 1]
    B = img[:, :, 2]
    hColorR = filter.hsobel(R)
    vColorR = filter.vsobel(R)
    hColorG = filter.hsobel(G)
    vColorG = filter.vsobel(G)
    hColorB = filter.hsobel(B)
    vColorB = filter.vsobel(B)

    energyArr = hColorR*hColorR+vColorR*vColorR+hColorG*hColorG+vColorG*vColorG+hColorB*hColorB+vColorB*vColorB

    return energyArr


def find_seam(img):

    height,width = img.shape[:2]
    seamFitness = dual_gradient_energy(img)

    #for i in range(0, width):
        #seamFitness[0][i] = img[0][i]
    for x in range(0, width-2):
        for y in range (1, height-2):
                #seamFitness[x][y] = img[x][y]
            if (x>0) and (x<width) and (y==0):
                seamFitness[x][y] += min(seamFitness[x][y-1], seamFitness[x+1][y-1])
                if (x>0) and (x == width-1):
                    seamFitness[x][y] += min(seamFitness[x][y-1], seamFitness[x-1][y-1])
                    if (x!=0):
                        seamFitness[x][y] += min(seamFitness[x-1][y-1], seamFitness[x][y-1], seamFitness[x+1][y-1])


    return seamFitness[y]


def remove_seam(img,seam):
    attempt = 0
    i = 0
    height,width = img.shape[:2]
    seamFitness = np.zeros((height, width))
    for attempt in range(attempt, img.size):
        bestRow = 0
        for i in range(i, height-img.size):
            if (seamFitness[width-1][bestRow] > seamFitness[width-1][i]):
                bestRow = i
    x = width-1
    if (x > 0):
        theMin = seamFitness[x-1][bestRow]
    if (bestRow > 0 and seamFitness[x-1][bestRow-1] <= theMin):
        bestRow = bestRow-1
    elif (bestRow < height-1 and seamFitness[x-1][bestRow+1] <= theMin):
        bestRow = bestRow+1
    return img

def plot_seam(img, seam):
    height,width,dim = img.shape
    for i in xrange(0,len(seam)):
        img[i][seam[i]][0] = 255
        img[i][seam[i]][1] = 0
        img[i][seam[i]][2] = 0
    #pass

def main():
    img = imread('someimage.png')
    img = img_as_float(img)
    l=dual_gradient_energy(img) #works!
    figure()
    gray()
    imshow(l)
    show()
    r = find_seam(img)
    print r
    s = remove_seam(img, r)
    imshow(plot_seam(img, r))
    show()

if __name__ == '__main__':
    main()

Open in new window

seamcarver.py
someimage.png
AttilaBAsked:
Who is Participating?
I wear a lot of hats...

"The solutions and answers provided on Experts Exchange have been extremely helpful to me over the last few years. I wear a lot of hats - Developer, Database Administrator, Help Desk, etc., so I know a lot of things but not a lot about one thing. Experts Exchange gives me answers from people who do know a lot about one thing, in a easy to use platform." -Todd S.

aikimarkCommented:
I would expect a floating point result, based on the definition of the energy gradient formula definition.  If you need integer results, cast the floating-point values into integer values.

I was curious why your x and y for loops have different starting values and (relative) different ending values.

btw...I'm a big fan of this algorithm and, generally, context sensitive resizing.
0

Experts Exchange Solution brought to you by

Your issues matter to us.

Facing a tech roadblock? Get the help and guidance you need from experienced professionals who care. Ask your question anytime, anywhere, with no hassle.

Start your 7-day free trial
It's more than this solution.Get answers and train to solve all your tech problems - anytime, anywhere.Try it for free Edge Out The Competitionfor your dream job with proven skills and certifications.Get started today Stand Outas the employee with proven skills.Start learning today for free Move Your Career Forwardwith certification training in the latest technologies.Start your trial today
Python

From novice to tech pro — start learning today.