The growing volume of high-resolution medical images demands efficient compression for storage and transmission. While JPEG is widely used, its lossy nature can introduce artifacts that compromise diagnostic integrity and hinder the performance of AI-driven analysis. This study evaluates VC-6, a novel,AI-friendly codec, as a superior alternative to JPEG for modern medical imaging workflows.

We performed a quantitative comparison of VC-6 and JPEG image

compression. Our source material was an endoscopy video encoded as a ProRes

file. For VC-6, TIFF frames were extracted and then encoded into the VC-6

format. For JPEG, the ProRes file was converted to MPEG4, and JPEG frames

were extracted at a quality level consistent with our standard workflow. We

then conducted a full-reference quality assessment using the decoded VC-6

image as the reference and the JPEG image as the distorted input. The key

metrics evaluated were Video Multimethod Assessment Fusion (VMAF), Contrast

Aware Multiscale Banding Index (CAMBI), Peak Signal-to-Noise Ratio (PSNR), and

Structural Similarity Index Measure (SSIM).

The results demonstrated a significant quality degradation in the JPEG

images when compared to the VC-6 reference. The mean VMAF score of 82.63 for

the JPEG images indicates a noticeable loss of perceptual quality compared to

the VC-6 source. The mean CAMBI score of 0.009 for the JPEGs suggests minimal

banding artifacts, but still measurable differences from the VC-6 reference.

The mean PSNR of 33.97 dB and mean SSIM of 0.9068 for the JPEGs further quantify

the distortion introduced by JPEG compression. In terms of file size, VC-6

files were found to be ~315% larger than JPEGs. However, both offer substantial

compression compared to uncompressed TIFFs, with VC-6 being 82.2% smaller and

JPEG being 95.7% smaller.

While VC-6 files are slightly larger than their JPEG counterparts,

this increased file size is directly correlated with superior image quality

and fidelity. More importantly, its inherent AI-friendly design, with its

multi-resolution representation, makes it an ideal format for next-generation

endoscopy AI video and image applications. By enabling more efficient and

accurate data access for machine learning models, VC-6 has the potential

to accelerate the development and adoption of AI-powered tools, where image

fidelity and AI utility outweigh the increased storage footprint.