In this work, I explore the themes of scale, convergence, and complexity by presenting decompositions of Edvard Munch's iconic painting “The Scream” at different resolutions. These images are composed of optimally sized rectangles, capturing the multiscale features of the painting. This is similar to the use of differently sized mesh elements to capture multiscale physics when solving PDEs. Further, while only 8 colors are used in these images, true colors can be recovered at a distance through convergence since the eye has limited resolving power. This series of images also provides a visual representation of an algorithm at work. Algorithms have become ubiquitous behind the scenes of modern life. Images are compressed, digitized, and distorted ever day without any thought of their aesthetic complexity. As we have discovered in the fields of sparsity and compressed sensing, compression reveals the fundamental structures in data. This work visually illustrates this idea. Process is a critical aspect of my work. These images were generated using MATLAB for decomposition. The algorithm segments the image into n rectangles by recursively subdividing them. The rectangle to be divided is chosen by maximizing the color separation between the two resultant rectangles, weighted by the size of the rectangles. After image segmentation, the color vector of the rectangle is decomposed into a black/while component, a RGB component, and two CYM components, which are able to span the full color spectrum without the need to mix any of the colors. Presented here are images composed of 1, 10, 100, 1K, 10K, 100K, and ~1M rectangles, where the final image uses every pixel as a rectangle.