Why JPEG Compression Converts RGB to YCbCr

The standard JPEG compression pipeline converts RGB color data into the YCbCr color space to exploit the physiological limitations of the human visual system. By separating an image's brightness (luminance) from its color information (chrominance), the JPEG algorithm can discard significant amounts of color data without causing a noticeable drop in perceived image quality. This separation enables techniques like chroma subsampling and targeted quantization, which are the primary drivers behind JPEG's ability to achieve high compression ratios while keeping file sizes small.

In the standard RGB model, red, green, and blue channels all carry both brightness and color information simultaneously. Because all three channels contribute heavily to the sharpness and structure of the image, discarding data from any single RGB channel introduces obvious visual artifacts and blurs detail.

Converting to YCbCr solves this problem by decoupling these components into three distinct channels:

• Y (Luminance): Represents the grayscale brightness and detail of the image.
• Cb (Chroma Blue): Represents the difference between blue and luminance.
• Cr (Chroma Red): Represents the difference between red and luminance.

The human eye contains significantly more rod cells (which detect light intensity and edge detail) than cone cells (which detect color). As a result, humans are far more sensitive to fine variations in brightness than to subtle shifts in color hues. By isolating luminance into the Y channel, JPEG preserves the details the human eye actually notices.

Once the data is in the YCbCr color space, the pipeline can apply chroma subsampling. This process downsamples the Cb and Cr channels to a lower resolution—commonly discarding up to 75% of the color data in a 4:2:0 subsampling scheme—while keeping the Y channel at full resolution. Because the luminance remains untouched, the human eye perceives the resulting image as sharp and detailed.

Finally, separating the channels allows for independent quantization after the Discrete Cosine Transform (DCT) stage. The JPEG algorithm uses different quantization matrices for luminance and chrominance, applying far more aggressive compression to the Cb and Cr channels. Without the initial conversion from RGB to YCbCr, these high-efficiency compression steps would not be possible.