Performance Analysis of Lossy Image Formats with Huffman Encoding Across Different Resolutions

This study compares the performance of four lossy image formats — HEIC, JPG, WebP, and AVIF — when combined with Huffman Encoding across different resolutions, evaluating image quality (PSNR, SSIM, MSE), processing time, and compression ratio to determine the most suitable format for different use cases.

Little research has thoroughly examined how Huffman Encoding affects image quality metrics when applied to various modern lossy formats, even as the need for efficient image compression keeps growing in domains such as medical imaging and real-time video streaming.

Three high-quality scenery photos (BMP format) were resized to five resolutions (640×426 to 5184×3456), processed with Huffman Encoding, then converted into four lossy formats (HEIC, JPG, WebP, AVIF). Each combination's PSNR, SSIM, MSE, processing time, and compression ratio were measured, repeated 30 times per resolution per format (450 total measurements).

Measurements were run on Google Colab using an AMD Ryzen 5 5500G CPU (6 cores, 12 threads); data normality was tested with Shapiro-Wilk, followed by the non-parametric Mann-Whitney U test since the data was not normally distributed.

No significant difference was found between formats in PSNR, SSIM, MSE, or compression ratio across all resolutions — but processing time differed significantly (Mann-Whitney test, p<0.001 for most format pairs). HEIC gave the highest PSNR/SSIM but was slowest (e.g. 53.5ms at the highest resolution); AVIF achieved the highest compression ratio (up to 372.5) while keeping reasonable quality; WebP sat in the middle, balancing quality and speed.

Provides practical guidance for choosing a lossy image format with Huffman Encoding: HEIC for quality-focused applications, AVIF for storage-constrained scenarios, and WebP as a balanced compromise for general needs.

The study used only three scenery images, did not account for different color spaces or varying levels of detail/texture, and tested only Huffman Encoding without comparing it to other lossless methods like LZW or RLE.