Application of Maximum Entropy-Based Image Resizing to Biomedical Imaging

Subsampling algorithms are applied to resize digital images to a lower resolution for display and transmission applications where the pixel count of the display mechanism is lower than the pixel count of the image acquisition method. Unfortunately, interpolation-based resizing methods change the color information and attenuate a specific range of high-frequency components from which the human visual system derives significant response. The described maximum entropy algorithm (MEA) provides that, as an image goes through subsampling, locally informative pixels are retained by analyzing the pixel neighboringhoods. The selected pixels are inserted directly in the output image, and color information is therefore preserved. From subjective observation and object evaluation using the entropy, contrast, and PSNR, MEA effectively maintains important features and color information and demonstrates better resizing performance than interpolation-based methods for some applications. Furthermore, the computational expense is suitable for real-time implementation