Split Image Compression: A Fun Exploration in Data Loss and Reconstruction
In this experiment, I developed a tool that splits an image into N vertical strips and uses an even-odd strip separation method to create two new images. Each of these images loses half the data from the original, but the overall shape and colors remain recognizable. The result is a lossy but understandable version of the image, as seen in the initial vertical cut of a chameleon image (see Figure 1).
Taking this experiment further, I applied the same method in the horizontal direction, splitting the two images again and generating a total of four images. With each step, more data is lost, yet the image is still identifiable (Figure 2).
If we increase the value of N (the number of vertical or horizontal strips), the quality of the output images improves, meaning that the image data is more spread out, and finer details are preserved. In theory, by reversing the process and combining the images, we could reconstruct the original image or at least achieve a very close approximation.
While this method is not the most practical approach for image compression due to its lossy nature, it was an enjoyable experiment to explore how image data can be manipulated, cut, and reassembled. It serves as a playful demonstration of how different techniques can be applied to compress and decompress image data, even if it seems counterintuitive at first.
You can use my site to do this experiment: https://imagecompress.brainbird.org
Conclusion: As you can see in some of the horizontally split images, there are noticeable gaps, and the strips don’t align perfectly. This misalignment occurs because we are using fixed-size squares to split the image horizontally. Since the image height may not always divide evenly into these square sections, the fit can be imperfect. Despite this limitation, the core concept of splitting and reassembling image data remains a fun experiment in understanding lossy image compression.
