Collagen image compression using the JPEG-based predictive lossless coding scheme

Abstract

Collagen second harmonic generation (SHG) images are helpful for the diagnosis of cancer and diabetes. In this paper, we develop the JPEG-based predictive lossless image coding (JPLIC) algorithm to encode SHG images. The main difference between collagen SHG images and other images is that they have very irregular texture and the correlations among adjacent pixels are very low. Therefore, conventional lossless compression algorithms, including JPEG-LS and CALIC, which are based on predicting current pixel values from neighbors, may not be suitable for collagen SHG image compression. Therefore, instead of applying adjacent pixels, the proposed algorithm predicts pixel values by JPEG. After prediction, we use adaptive arithmetic coding together with context modeling to encode the residue. Simulations show that the proposed JPLIC algorithm has better performance than other lossless compression methods for collagen SHG images. Moreover, the proposed algorithm is also suitable for nearly lossless image compression and noise-like image compression.