Coarse-Grained
This method simplifies the image by dividing it into blocks and replacing every pixel within a block with that block's average color. The slider below moves from most coarse (left) to most granular (right).
Quantization
This technique reduces the number of distinct colors in an image. Each pixel's Red, Green, and Blue component is a value from 0-255 (8 bits). Quantization maps these to a smaller set of levels, reducing the data needed.
Block Polynomial Fit
This method breaks the image into small 8x8 pixel blocks. For each block, it calculates a mathematical polynomial function that best approximates the color values. Instead of storing every pixel, it only stores the function's coefficients, saving space. The "Polynomial Degree" determines the function's complexity.
Discrete Cosine Transform
This is the core method used in JPEG compression. It transforms 8x8 blocks of pixels into a frequency representation using a Discrete Cosine Transform (DCT). It then discards high-frequency details, which are less noticeable to the human eye, keeping only the most important "coefficients" to reconstruct the image.