pixel-blending

Pixel Blending

Challenge

This is a simple pixel blending challenge which could benefit from multithreading, SIMD and even GPGPU techniques, such as CUDA.

The challenge is to take an image containing a number of sprites (i.e a spritesheet) and to blend the pixels in those sprites following the algorithm described below.

Algorithm

Calculate the final pixel values by doing the following:

• Declare a variable "result" which is a 2D array of 3D vector. Initialise it with the size a x c and all zero values. This array will contain the final pixel outputs for the output file.
• Iterate over all n sprites in the spritesheet.
• Within each loop, update the "result" by doing the following calculation:

result.r = result.r * 0.5 + subimage[index].r * 0.5
result.g = result.g * 0.5 + subimage[index].g * 0.5
result.b = result.b * 0.5 + subimage[index].b * 0.5

• where index is the loop iteration number
• subimage[index] refers to the sub-image within the input file where the pixel starts from x: 0 and y: "c" x "index".
• r, g and b represent the corresponding color values.

Dataset

Input file

1. Input file is an image file in .PPM (Portable Pix Map) extension. More info on PPM here - https://www.cs.swarthmore.edu/~soni/cs35/f13/Labs/extras/01/ppm_info.html

• For this challenge, we will restrict the first header line to be "P3" and the max color range to be 255.

2. The image file is of size a x b. (a is the column pixel count and b is the row pixel count)
3. The image can be thought of as a "spritesheet", a list of images of size a x c combined into one. (where "b" is a multiple of "c")

Example:

Output file

1. Output file will also be an image file in PPM format.
2. The image file should be of size a x c (where "b" is dividable by "c" )

Credits

Thanks to Jessub Kim, who originally proposed this challenge here.