Pathological microscopic image compression using lossless and progressive coding schemes

TLDR: The proposed lossless Karhunen-Loeve Transform based on ladder networks is proposed, which reduces inter-color redundancies which increases coding performance and progressive transmission capability and the coding performance is superior to that of DCT-based JPEG with the RGB/YUV transform.

Abstract: This paper proposes pathological microscopic image compression schemes that suit lossless and progressive transmission. Because pathological microscopic images require very high resolution, they create heavy storage requirements and long transmission times. Image compression is desired to reduce these problems. First, we propose a lossless Karhunen-Loeve Transform (KLT) based on ladder networks. The proposed lossless KLT reduces inter-color redundancies which increases coding performance. Next we propose a progressive transmission algorithm by combining the lossless KLT and set partitioning in hierarchical trees (SPIHT) with the S+P transform. SPIHT is adopted to encode individual color-transformed components. By considering coding efficiency, the transmission bit rates of each encoded component are determined. The resulting algorithm gives high coding performance and has progressive transmission capability. When all transmitted data are decoded, decoding yields the original image. We demonstrate the performance of the proposed algorithm when applied to super high definition pathological microscopic images. All the images used in our tests have 2048x2048 pixels and 24 bits per pixel. It is shown that the coding performance of the proposed algorithm is superior to that of DCT-based JPEG with the RGB/YUV transform.