ABSTRACT
Confocal scanning microscopy is a commonly used method for acquisition of high-quality digital two-and three-dimensional images of molecular biological objects. The high quality of confocal images makes it possible to extract quantitative data at a single cell resolution, the availability of which is a necessary prerequisite for successful systems biology studies. However the data accuracy is limited due to errors that arise in the course of confocal scanning. In our recent papers [1,2] we analyzed the sources of errors introduced by two-dimensional confocal imaging into the data on gene expression in situ and described algorithms for estimation and correction of these errors. For example, confocal images are inevitably
contaminated by photon shot noise [3] and a common way to reduce the noise is the averaging of multiple separate scans. However, the information about the averaged image will be lost if pixels with high or/and low intensities are clipped in single scans. Image clipping is a form of signal distortion related to the limited grayscale range of an image. Pixel values that exceed an upper threshold of the grayscale range (e.g., 255 for an 8-bit format) are cut-off at the threshold value, all the pixels with negative intensities are set to zero. Such pixels are referred to as over-and under-saturated, respectively. Averaging of clipped scans results in errors in the data extracted from the averaged image. In our previous work we developed a method [1] based on censoring technique for estimation and correction of this kind of errors, however the method implementation requires not only the averaged image but also all the confocal scans which are not provided by the standard procedure of image acquisition.
