Hybrid Region-based Image Compression Scheme for Mamograms and Ultrasound Images

The need for transmission and archive of mammograms and ultrasound Images has
dramatically increased in tele-healthcare applications. Such images require large
amount of' storage space which affect transmission speed. Therefore an effective
compression scheme is essential. Compression of these images. in general. laces a
great challenge to compromise between the higher compression ratio and the relevant
diagnostic information. Out of the many studied compression schemes. lossless
.
IPl. (i-
LS and lossy SPII IT are found to he the most efficient ones. JPEG-LS and SI'll IT are
chosen based on a comprehensive experimental study carried on a large number of
mammograms and ultrasound images of different sizes and texture. The lossless
schemes are evaluated based on the compression ratio and compression speed. The
distortion in the image quality which is introduced by lossy methods evaluated based
on objective criteria using Mean Square Error (MSE) and Peak signal to Noise Ratio
(PSNR). It is found that lossless compression can achieve a modest compression ratio
2: 1 - 4: 1. bossy compression schemes can achieve higher compression ratios than
lossless ones but at the price of the image quality which may impede diagnostic
conclusions. In this work, a new compression approach called Ilvbrid Region-based Image
Compression Scheme (IIYRICS) has been proposed for the mammograms and
ultrasound images to achieve higher compression ratios without compromising the
diagnostic quality. In I LYRICS, a modification for JPI; G-LS is introduced to encode
the arbitrary shaped disease affected regions. Then Shape adaptive SPIT IT is applied
on the remaining non region of interest. The results clearly show that this hybrid
strategy can yield high compression ratios with perfect reconstruction of diagnostic
relevant regions, achieving high speed transmission and less storage requirement. For
the sample images considered in our experiment, the compression ratio increases
approximately ten times. However, this increase depends upon the size of the region
of interest chosen. It is also föund that the pre-processing (contrast stretching) of
region of interest improves compression ratios on mammograms but not on ultrasound
images.