Intensity based multi-scale blending for panoramic images in fluorescence endoscopy

Panoramic images providing a larger local overview of the internal urinary bladder wall can be used for documentation and surgery planning, as well as assist the re-identification of multi-focal tumors during a cystoscopy. In contrast to white light illumination, the photodynamic diagnosis (PDD) leads to an enhanced tissue contrast, resulting in a reddish fluorescence of malignant tissue excited by a bluish small-band illumination. Due to the low illumination power, the fluorescence intensity and the contrast of vasculature are heavily dependent on the distance between the endoscope and the bladder wall. Thus, images with varying illumination and different resolved structures are combined during the iterative panorama composition. In this case, common linear blending methods applied by mosaicking algorithm for endoscopic images, generate strong visual interpolation artifacts in the output image. Instead, the developed intensity based multi-scale blending method, tailored to fluorescence endoscope images, provides a better visual panoramic image quality. Based on a highest intensity decision, a non-linear weighting function is applied on several sub-bands of a Laplacian pyramid image representation. Thus, bright fluorescence information and small vessel texture are preserved in the image composition at the same time. Besides an image quality improvement, the blending algorithm permits a more comfortable and unrestricted endoscope movement for the physician during the bladder scan.