Pathology Image Compression Based on JPEG2000, Multi-Resolutional Human Perception and the Region of Interest Predictions

Abstract

To achieve high efficiency of remote pathology image browsing in telemedicine, efficient image compression coding is required. In this work, we establish a visibility threshold (VT) model, which considers multi-resolution and different visual qualities jointly. Based on this model, we propose an image coding method, which operates adaptively according to the required resolutions and visual qualities, under the JPEG2000 standard for the whole-slide pathology images (WSIs). Particularly, we calculated the average standard deviation of the wavelet coefficient from lung squamous cell carcinoma (LSCC) WSIs and implemented in the VT calculations, as well as the masking factor calculation in each wavelet subband to count for the masking effect. The objective metric, HDR-VDP 3.0.6 score was implemented to evaluate to proposed encoding method. The averaged resulting scores across 100 LSCC WSIs in various visual qualities are plotted in Fig. 1, where half of the full resolution was used. The results from the mean-squared-error (MSE) based encoder at the same average bit rates are also shown. It can be seen that the predicted visual quality of decoded image from the proposed method is improved, compared with the MSE based method. Additionally, we also propose a weighting model to adjust the VTs. The encoder with the adjusted VTs concentrates on retaining the visual quality for the regions, where the lesion probabilities predicted with deep neural networks are high. A representative comparison of the decoded WSI blocks from the encoders with the adjusted and the non-adjusted VTs is shown in Fig. 2, where the adjusted VTs are shown to retain important features in decoding results while further reduce the coding bit rates.