First, the original image is smoothed by a 5x5 Gaussian filter. This pre-processing step helps reduce the noise components prior to what I wanted to do next. The kernel I used for this step is: 1 2 4 2 1 2 4 8 4 2 4 8 16 8 4 2 4 8 4 2 1 2 4 2 1 Then the image is normalized to the range 0-255 by divisor 100, which is just the sum of all the weights. Next, I took the resulting image and applied a 3x3 Laplacian-like filter to it. How is it "Laplacian-like"? Most of the time, the coefficients of a Laplacian filter would sum to 0 and hence no further normalization is required. One of the most common Laplacian filter kernels is: -1 -1 -1 -1 8 -1 -1 -1 -1 The Laplacian of an image serves to highlight regions of rapid brightness change and is therefore helpful for sharpening the contrast. The above filter sharpens a response value by giving it more weight (in this case, 8) and pulling down its neighboring values. However, being amazed by the color contrast in which it results, in my artifact I intentionally changed the middle 8 to a 4, thereby giving the image a more artistic and impressionistic sense (from my point of view). Finally, as a wrapping up step, I further smoothed the image by applying a 3x3 mean filter to it.