Abstract
Purpose::
The need for multiple digital color images for clinical research and telemedicine remote evaluation of diabetic retinopathy can require significant computer storage and high transmission speeds. Joint Photographic Experts Group Compression (JPEG) algorithms are widely used techniques to lower computer file storage and speed transmission requirements, though potentially at some "lossy" degradation to image fidelity. Both JPEG & JPEG 2000 allow selection of compression from a range of quality factor options. This study correlates image fidelity of JPEG and JPEG 2000's range of quality factors.
Methods::
One hundred fifty-one 3008 x 2000 pixel diabetic retinal images were compressed using JPEG and JPEG 2000 algorithms. Each image was compressed using a quality factor of 100%, 95%, 85%, 80%, 65%, 50% and 1%. Just noticeable differences (JNDs) between compressed images were measured and compared to uncompressed images using Color JND-Metrix computational observer software.
Results::
Image fidelity between 0.1 and 0.01 JND was achieved with quality factors above 70% using either JPEG or JPEG 2000. At the same quality factor, JPEG 2000 achieved higher compression ratios than JPEG. The highest image fidelity was produced with JPEG 2000 at 100% and JPEG at 90%, achieving a compression ratio of 37:1 and 28:1 respectively.
Conclusions::
JPEG and JPEG 2000 can both preserve image fidelity but JPEG 2000 achieves a higher compression ratio.
Keywords: diabetic retinopathy • imaging/image analysis: clinical • retina