With the widespread utilization of charge-coupled-devices, there is much interest in methods to efficiently process images. The processing, manipulation, and storage of photographic quality digital images place significant demands on today's computers. Even with today's high performance bus structure and real-time operating systems, manipulating full resolution image data may quickly overwhelm computer hardware and software. In response to this, data reduction techniques have been developed to aid in resolving this problem. Two common data reduction techniques include data sub-sampling and data averaging. Data sub-sampling approach is simplistic in nature and perhaps easiest to implement in both hardware and/or software. This approach involves sub-sampling the full resolution image data to a lower resolution. Selection of sub-sampled element of the full resolution image is random in nature. This random selection makes sub-sampling an effective technique for flat image fields but degrades or softens the image for edges information quality/content. Data averaging approach is more difficult and complex to implement in both hardware and software than the sub-sampling approach. The data averaging approach involves a two dimensional averaging function to sub-sample the full resolution image data to a lower resolution. Averaging area parameters may be chosen to average X consecutive pixels, and Y consecutive lines. Although more complex, data averaging more effectively retains edge information. This thesis investigates the two-dimensional, pixel data-averaging method for data reduction. It supports the use of a pixel-averaging algorithm in conjunction with, or independent from compression techniques which may be employed elsewhere within the same system. Hardware and software implementations are presented to solve this system problem. The hardware architecture design is based on a pixel averaging application specific integrated circuit. Software routines written in C programming language are presented to perform this data-averaging task. Performance comparisons are made between the hardware and software implementations for image resolutions up to 2048 by 3072 pixels, and under several averaging conditions. This thesis also provides a survey of various types of charge-coupled devices sensors, focusing on their abilities and limitations for data averaging. It presents several applications where this type of data reduction would be advantageous.
Library of Congress Subject Headings
Image processing--Digital techniques; Data reduction--Computer programs
Department, Program, or Center
Computer Engineering (KGCOE)
DeSanctis, Paul, "Analysis of H/W & S/W techniques for data reduction in high speed digital image processing" (2002). Thesis. Rochester Institute of Technology. Accessed from
RIT – Main Campus