Gray-scale resolution enhancement

The number of bits assigned to represent the color intensity at image pixels is usually referred as the bit depth. When the bit depth is not sufficient, images suffer from ridge-like structures known as false contours. Bit-depth limitations become important when low-contrast details are required, as in medical imaging, aerial/satellite photography, and high-quality scanning applications. In this paper, we investigate a method for increasing bit depth. Specifically, we show that when a sequence of video frames is available, then it is possible to achieve a higher bit depth through a projections onto convex sets (POCS) based reconstruction method. INTRODUCTION When images are digitized, a certain number of bits is assigned to each pixel to represent its intensity. The number of bits, the bit depth, determines the number of gray levels between the minimum and maximum intensities that the imaging device can capture. There will be a loss of gray-scale resolution if the bit depth is not sufficient. When a set of low bit-depth images is available that are slightly different from each other because of motion or illumination, their non-redundant information can be combined to enhance the gray-scale resolution. We refer to this process of multi-frame gray-scale resolution enhancement as superprecision. Superprecision can be used in several application areas. One of the most important of these is medical imaging. In medical imaging low-contrast details are often extremely critical for diagnosis, but when the bit depth is insufficient, these details may be lost. Superprecision reconstruction has the potential to regain these details by combining the non-redundant information that is present in a set of images. Military automatic target detection, aerial and satellite remote sensing, and high-quality scanning applications can also make effective use of superprecision reconstruction to enhance gray-scale resolution. The conversion of images from a low bit-depth format to a higher one (e.g., conversion from 8-bit GIF to 24-bit JPEG) is also a potential application of this technology. The superprecision problem is similar to the superresolution problem where higher spatial resolution is sought from a set of low spatial resolution images [1]. Although the superresolution has received a significant amount of attention, the superprecision has not been researched sufficiently. In an early paper, Cheeseman [2] proposed to increase the spatial and gray-scale resolution at the same time by using a Maximum probability approach. They assumed a Gaussian model for all This work was supported by the Office of Naval Research (ONR) under the award N000140110619. 1 ( ) 1 2 ( , , ) N g m m k 1 ( ) 1 2 ( , ) N f n n Image capture Bit-depth reduction 2 ( ) 1 2 ( , , ) N g m m k