Back
Close
  • 86

Learning Opportunities

This puzzle can be solved using the following concepts. Practice using these concepts and improve your skills.

Statement

 Goal

Seam carving is a content-aware image resizing technique that often leads to impressive results despite being very simple. When used to reduce the size of an image, it intends to remove the supposedly least significant parts of the image instead of resizing everything in a uniform manner.
In this problem, we implement the main algorithm behind seam carving. To focus on the algorithmic aspects, we only reduce the width of grayscale images.

Given a grayscale image of size W×H, let I(x,y) be the 1-byte intensity (from 0 for black to 255 for white) of the pixel at (x,y) with 0 ≤ x < W and 0 ≤ y < H; (0,0) being the top-left corner of the image. We define the intensity differentials and the energy function as:
dI/dx(x,y) = I(x+1,y) - I(x-1,y) if 0 < x < W-1
0 otherwise (left/right borders)
dI/dy(x,y) = I(x,y+1) - I(x,y-1) if 0 < y < H-1
0 otherwise (top/bottom borders)
E(x,y) = |dI/dx(x,y)| + |dI/dy(x,y)|
(where |.| denotes the absolute value)

A vertical path is a sequence of pixels (x(0),0), (x(1),1), ..., (x(H-1),H-1) such that |x(i+1) - x(i)|≤1. In other words, the path contains exactly one pixel per line and consecutive pixels are vertically or diagonally neighbors. The energy of a path is the sum of the energies of its pixels.

To reduce the width of the image by one pixel, compute a vertical path of lowest energy and simply remove it.
To reduce by several pixels, simply repeat that 1-px reduction step on the successive reduced versions of the image.

Instructions
Given a PGM image (see input description) and its desired new width, resize it using seam carving.
When there are several paths of lowest energy, remove the lexicographically smallest one over the (x(0), x(1), ...) values (i.e. the leftmost path from top to bottom).
Due to the limitations on the output size, you are not asked to return the resulting image but the energies of the successively deleted paths.

Additional remarks
Try it on your own images and visualize the results! You might have to relax your PGM parser as the specifications of the format do not actually make the space/newline distinction. For very large images, make sure that you use an efficient algorithm and optimize your code so that it keeps and updates the intermediate results after each path deletion (to avoid of recomputing everything at each step).

Check out the references for more information (alternative energy functions, strategies to resize both horizontally and vertically, use seam carving to enlarge images, etc).

References
https://www.youtube.com/watch?v=6NcIJXTlugc
http://www.faculty.idc.ac.il/arik/SCWeb/imret/imret.pdf
Input
The input is a valid PGM-ASCII file.
Line 1 always contains the magic number P2 (indicating an ASCII graymap).
Line 2: Two space-separated integers W and H corresponding to the width and height of the image.
Line 3: The character # (to start a comment line in the PGM file) followed by a space and an integer V indicating the target width.
Line 4 always contains 255, indicating that the intensity is coded on one byte by an integer between 0 and 255.
Next H lines: W space-separated integers between 0 and 255 corresponding to the intensities of the pixels of the successive lines of the image (from top to bottom).

See also https://en.wikipedia.org/wiki/Netpbm_format for more information about this format.
Output
W-V lines indicating the energies of the successively deleted paths.
(You are not asked to output the resulting PGM file due to output sizes limitations.)
Constraints
5 ≤ W, H < 100
3 ≤ V < W
Example
Input
P2
30 8
# 29
255
164 174 175 173 176 171 171 169 165 162 164 166 165 162 168 169 163 174 207 173 187 196 183 177 178 164 158 162 168 185
179 130 118 108 117 136 148 165 165 162 155 156 157 155 156 162 170 161 134 100 109 111 117 132 140 154 151 139 125 136
93 74 102 87 73 98 98 92 96 137 146 138 136 134 132 131 133 128 91 110 112 104 78 77 86 103 122 147 150 146
129 140 152 125 59 62 82 76 54 107 156 145 139 141 145 142 138 138 141 144 146 149 90 64 84 93 70 92 177 174
167 159 160 133 101 83 86 59 29 67 167 171 169 155 147 142 136 129 129 126 141 113 90 74 54 71 46 51 163 189
192 198 157 160 170 134 195 114 156 109 114 197 200 199 193 183 179 168 156 163 127 126 161 122 103 131 75 88 96 183
190 198 153 170 173 138 129 150 191 197 118 167 202 202 204 194 187 185 184 142 132 172 171 150 125 161 135 190 145 132
188 188 175 173 154 142 141 170 161 154 116 134 154 160 176 192 188 184 185 159 161 158 162 128 118 121 147 159 171 111
Output
175

A higher resolution is required to access the IDE